community-help

[fossil jolt]

How to make codex - either CLI version or VS code version to always write the files without my confirmation?

[analog herald]

Hey anyone here running into bad voice issues on Team plan since last friday?
I am in the second month on a Team plan.
After migration all went fine.
Now it is a week since

• voice mode has clack clack when answers
• thinking is around 2 to 3 minutes IF it responds but often does not just does a clack and over
• asking even short questions does not transcribe so the AI does not answer.
• in perfect silence no talks around no noises the sistem sometimes hear things and type them like things about video or thank you or other weird stuff.
• GPT 5.0 Memory issue - when I cleaned the memory once and tried to add other memories it just defaulted to a placeholder and refused to update memory. I had to update memory entries using gpt4.0 in small batches.
• When spoken not long questions or talked normally there is one week since it is not functional in minimal parameters namely the voice is either ignored or transcribed in some weird things about videos leaving both teams without understanding what is it about.

do not tell me to talk to the assistant help… did and… honestly I can even talk in the mirror. It cannot do much

many thanks for any kinfds help

I tried android app, browsers( all of them) windows app.

[twilit barn]

Hey, whenever I try to use Codex in VSC, I get error "Error starting conversation" or "Error starting cloud task" (depending on whether I run it locally or in the cloud). This happens for any prompt in any project. Anyone know what could be causing this? I've already tried reinstalling the plugin and relogging.

[compact sage]

I’m on the Plus plan, so why can’t I generate images in Sora anymore? It keeps showing me the message: ‘You can only generate 1 image or video at a time. Upgrade to generate more images or videos at the same time.’ But I wasn’t even generating anything when it popped up. And now it shows up all the time.

[undone charm]

guys, why cursor charged me twice? I am subscribed $20 per month every 11th of the month, but why they charged me in aug 8 with 16.86?

...cursor said "Pro comes with at least $20 of included usage per month. ". Why?

[sleek lagoon]

@Hueseyin I suggest using ChatGPT Exporter. It can export to variety of formats which are free (PDF export is limited to 3 a day), but you can easily export to markdown or html (I use exclusively markdown) and then use either MD reading tools or plugins - I use Markdown viewer,to display it and print in the format you wish. Cannot provide URLs here due to restrictions but for the first one the "fingerprint" is: Version
2.26.0
Updated
August 25, 2025
Size
954KiB
Languages
34 languages

Hopefully you find the exact one.

Too bad I cannot give exact URLs, because there are many fakes/scams out there, but hopefully you find the right one. For the Markdown viewer I use the following chrome plugin.

Version
5.3
Updated
2 May 2024
Size
2.36MiB
Languages
English

[trail cairn]

cursor has its own discord, its not really openai related

[lone cargo]

How to use the new GPT-realtime??

[charred parcel]

do you mean the Realtime API?

[lone cargo]

Yeah

in the last openai live

[silk glade]

I keep getting a play verification error and chat gpt won't let me login it's been like this for maybe four days.

I get this error constantly even uninstalled and reinstalled.

[charred parcel]

here -> https://platform.openai.com/docs/guides/realtime#page-top

[lone cargo]

Thank you!

[lethal thicket]

Sorry for cross posting here.... I'm trying to use the Notion MCP connector, and it keeps stating that it doesn't have write permissions, when it clearly does. Using Warp, everything just works. When pressed, GPT5 in chat claims that while it sees it has write permissions, OpenAI doesn't allow the writing function. Can anyone confirm that is the expected behavior, and why?

[trail cairn]

chatgpt mcp is only for retrieval

[zenith mist]

Hey, I am getting this message that image/video we are trying to generate is against the terms of service... Is there a way to talk to someone privately about it?

[trail cairn]

nope

[late atlas]

u want to ask why is going against ToS or to ask for 'bypass'?

[rustic marlin]

unless you have some clearly problematic prompt, it may be possible it is a false positive, share your prompt (if it does not break the rules) on #sora-discussions and ask for help about how to word it better

[short belfry]

Does anyone know why gpt-realtime doesn't work?

main.ts:

import { RealtimeAgent, RealtimeSession, OpenAIRealtimeWebRTC } from '@openai/agents/realtime';

const agent = new RealtimeAgent({
  name: 'Assistant',
  instructions: 'Translate video to Russian in realtime.',
});

const session = new RealtimeSession(agent, {
  transport: new OpenAIRealtimeWebRTC({
    useInsecureApiKey: true,
  }),
});

await session.connect({
  apiKey: 'sk-proj-...'
});

[formal meadow]

For some reason, When I ask ChatGPT 5 to use it's Thinking Mini or Thinking Model it just doesn't think for some reason and It didn't do this before?

[timber shadow]

I'm not familiar with that email address personally. Per this article: https://help.openai.com/en/articles/6614161-how-can-i-contact-support the official way to reach out to OpenAI support is via the chat widget on help.openai.com (icon in the bottom-right).

[timber shadow]

The domain does seem to be listed in this article: https://help.openai.com/en/articles/11725090-verifying-communications-from-openai#h_31633229a3 though not that specific address. So it is possibly a real OpenAI email address. If you haven't already tried sending your inquiry to the chat widget on help.openai.com though, could be worth it to start there, as that's the recommended path for reaching out to support (according to the first link I shared)

[tender badge]

0 0.235

[celest cradle]

I automated some image generation in order to waste $50 so I could get to tier 2 so I could see what the limits are at tier 2 because I want to try to use gpt-5 w/ my Cline agent. How long do I have to wait for them to upgrade me to tier 2? Does that happen daily or something?

[tawdry timber]

im confused is codex just messing up for me or did using the desktop version make my normal codex on web have to wait 5 or 7 days whatever it said i forgot

[lethal gust]

How do I get this black box to appear because I’m trying to automate a complex task but it isn’t generating the box you see in the screenshot!!

[timber shadow]

That black box looks like the UI for the tasks feature. I find that works best if I phrase the query like "Every day at x, give me y" or "Remind me to do x at y o'clock." So, it may be a phrasing thing regarding why it's not working for more complex tasks. Could you share a little more about what isn't working?

[lethal gust]

Yeah sure. So every time my ChatGPT app updates I always ask it if OpenAI or whoever works on update have implemented this system

“Feature
Status
Universal folder visibility
🟢 Released (July 2025)
Non-project chats visibility
🟢 Bug fixed (August 2025)
Delete individual messages
⚪ Requested
Nested folders / sub-projects
⚪ Requested
Pinning chats/folders
⚪ Requested
Color coding / tags
⚪ Requested
Multi-user live collaboration
⚪ Requested
Video parsing (e.g., YouTube)
⚪ Requested
Project dashboards / linking
⚪ Requested
Offline draft mode
⚪ Requested
Unified All Chats View
🟢 Functionally achieved” but I’m wanting it to be constantly tracked every lunch time UK time and I was half expecting the box to show up again like the previous screenshot showed but this screenshot I’m sending shows that it doesn’t always show up like that but it’s claiming that the automation has been set up but I’m very suspicious that it may not actually be set up.

It’s been talking about, I can set up manual automated tasks as well but I can’t find it anywhere on my IOS app

[timber shadow]

Hmm, hard to say exactly without seeing the phrasing of the actual task request, but the basic idea should work.

It might be all the added complexity of the specific things in your request throwing it off

[lethal gust]

“Set up an automation that runs every day at 12 PM UK time. Have it check if OpenAI/ChatGPT has updated or announced new features like delete individual messages, nested folders/sub-projects, pinning chats/folders, color coding/tags, multi-user live collaboration, video parsing (e.g. YouTube), project dashboards/linking, offline draft mode, or unified All Chats view — and tell me in this chat what’s changed.” Here, here’s my prompt

[timber shadow]

Do you have existing scheduled tasks already? You can check here: https://chatgpt.com/schedules

Maybe try omitting "UK time" too, it should localize automatically

[lethal gust]

How do I access that in the IOS app version?! 😱😱😱

[timber shadow]

I'm not sure if it's in the apps currently! But it's accessible in the browser. I'm on Android personally

[lethal gust]

Is it in MacOS or Web?

[timber shadow]

Not sure on the Mac app either! But definitely on chatgpt.com

[lethal gust]

Right! What about manually making automated tasks like it claims?!

Not via prompts I mean, it was talking about there’s a button that means that you click on it and get to fill out a form for an automated task.

[timber shadow]

Try creating a simple task in the browser version, then click these three dots

Then edit

You'll see some manual edit fields after that!

[lethal gust]

That’ll be the only way to do it then.

[charred parcel]

it needs to be noted that ChatGPT does not reliably know it's own capabilities, so prompts like yours may lead to inaccuracies in response.

[lethal gust]

I just need to understand that the black box won’t always appear for me.

[timber shadow]

Asking for the model to search online for sources for the task will probably return the best results, but yes pure self-knowledge is not something ChatGPT should be used for much

[lethal gust]

It kinda does need to know that in my opinion hahahahahahahaha. I’m just saying that out of respect because it’s making me feel like sometimes it can do anything

[lethal gust]

Fair enough

[timber shadow]

That's why they always have the disclaimer about being wrong sometimes 🙂

[lethal gust]

Very fair!

[late atlas]

@Kalin you have as model gpt-4o.. for tasks to work you need o3 and up (so now gpt-5)

else if you have other model that dont support task, it will just hallucinate about it.. tell u that will do it etc..

Also on the task prompt, it may help to have the actual url from openai changelogs (which is where u need to look up)

[lethal gust]

That’s what I was trying to figure out!!!! I was wondering why the black border box wasn’t showing up when I was creating my new automated tasks and I wasn’t sure if I was on the right model!!

[lethal gust]

What do you mean?

[late atlas]

https://help.openai.com/en/articles/6825453-chatgpt-release-notes

tell it to look it up here (if is only chatgpt u care)

but the way u have it also will work cause is something easy to search

i have similar task, bringing daily news on ai about specific things

[lethal gust]

I put the release notes on the edited task.

[late atlas]

ye.. as url and just ask to look up there for the XYZ things u want

nice features btw that you wait for xd

[lethal gust]

Oh really? Thank you so much

[lethal gust]

I appreciate that

[late atlas]

cheers! did it work?

[lethal gust]

I’ll see at lunch

[mighty field]

can someone help please?((

[celest cradle]

I'm stuck at tier 1 for a month now.

[exotic nymph]

Any ideas how I get rid of a video generation that has been sat with just a spinning circle for the last day?
I’ve tried deleting it but it keeps coming back. I’ve cleared my cache but nothing is getting rid of it. Any ideas?

[cinder harness]

Hi all

I was using ChatGPT via Windows app and it started being incredibly show to input text. I then started a GPT window on Chrome browser and it's the same. This seems to be happening since GPT5 went live. Is this a common issue? Would love to know how I can get back to it running smoothly. It's really disrupting me. Copilot is soooooo smooth compared to this but to me isn't as good for what i need.

Appreciate if anyone can help.

[silk glade]

Where do I get help if I get a play verification error? I've deleted cache etc. So another AI tells me I'm actually soft banned. The reason would be nuts for asking it to "write an antigenocide anthem."

[exotic nymph]

I am finding GPT5 to be extremely slow on all platforms, so much so that I’ve gone back to 4 until they improve performance.

[lethal thicket]

hmm, That seems a bit weird, since other MCP cliens, like Warp, are happy to write to it.

[trail cairn]

https://platform.openai.com/docs/mcp#create-an-mcp-server

[winter cobalt]

Anyone else notice that GPT-4o has changed to worse again? It doesn't even remember the second-to-last message in a new chat, it seems to be a completely different model.

[vivid mortar]

Hmm I have tried to replicate this and I cant seem to. You may just need to start a new chat :)

[rugged wing]

What do I do if I’m getting a unusual activity has been detected from your device error message

[cinder harness]

Yeah I’m thinking to do the same. On another forum someone suggested it’s due to it being a long conversation. I’m relatively new to this stuff in the last 6 months and I didn’t know a long conversation can make it slower. Will try and move to a new conversation but I’m worried about it not seeing the detail on the previous one

[exotic nymph]

Long conversations do slow down every model. If I know it’s going to be long, or quite detailed then I start a project so I can have multiple conversations

[wary agate]

Hey folks, quick question, does OpenAI (or anywhere legit) offer a free trial or promo for ChatGPT Plus right now? Not looking for anyone’s account or anything, just curious if there’s an official way to test it out before subscribing. Also open to hearing if there are other AI tools with free trials worth trying. Thanks

[charred parcel]

the entry level way to test ChatGPT is to use it on a Free account

[wary agate]

correct, but it limits on how much i do on it, hence asking the question

[charred parcel]

there isn't a free trial for ChatGPT. There are some other providers who have the GPT models available but that's not the same as ChatGPT

[wary agate]

no problem, thank you, was just asking because i saw on the site that some users have referral link invites for plus. Dont mean to cause any harm, thanks for the reply

[mystic nova]

I noticed too. It's as though reference chat history has been turned off. Like my chat has been lobotomized.

[tropic ocean]

Guys my Custom GPT sometimes forgot the system prompt,any fix?

[supple hatch]

Sounds like context needs refreshing

[summer grail]

Hi, I have a question. I'm testing the "thinking" model further, and I'm running into a problem. Whenever I ask something, it gives a mini-answer and then a mega-action plan based on it. I mean, it doesn't bother to be precise in its response. To make up an example, if I ask it "Why is it important to go to the gym when exercising?" and it would most likely tell you "Because it's healthy, here's a routine and a meal plan for a whole year..." Is this a problem with Prompto? Is this a problem with the model? Because it's also very uncomfortable having to be super-explanatory and super-specific so that the Thinking model responds precisely to what I need.

[ancient herald]

Did you guys know Pro have a limit on Agents use?

also I'd be curious to know what happens when my pro subscription renews on September 11th. Will the limits reset or do I drag them till September 22nd anyway? What if I cancel and resubscribe?

[supple hatch]

Yes, it has monthly limits that don't carry over. 300 runs. It would be significantly heavy use if you're using Agent 10 times every day!

Cancelling and resubscribing to exploit it would be a breach of terms of service

[trail cairn]

cancel and resub does not reset anything unless you resub to a higher tier which is not possible with pro

[cosmic mantle]

Every time I try to report a bug I get this step but this procedure literally does not exist on the app? It just takes me to Help Center and brings up that browser pop up dialogue. Am I missing something here? On the latest chatgpt app version

[supple hatch]

When you figure out how, report this bug too 😂

[pastel crow]

Has anyone else noticed that the quizgpt mode in ChatGPT hasn’t been working today?

[late atlas]

what is quizgpt mode?

[supple hatch]

Yeah - I mentioned it as a tip to someone in here how was asking how to use chatGPT to learn more. Then tried it and it didn't work!

Wonder if it's a feature that might come out and we weren't supposed to access yet.

[pastel crow]

It’s an interesting thought, and it may well be that OpenAI accidentally launched the feature by mistake. At the same time, we remember that they announced the rollout of the QuizGPT mode on their own Instagram, and the post is still up where they present the feature as available to everyone. It has also worked for me before. Has anyone else gotten it to work recently, and if so, more reliably than I have? I really hope they haven’t removed it, because it was a great feature for learning in a creative way.

[inland apex]

Does OpenAI post a known problems kind of thing?

[trail ingot]

Hey all! I'm new here and new to ChatGPT. I was having it help design lesson plans for a preschool and a few things it said it did was make supplemental materials such as coloring sheets and templates for games but when I open the file, theres just words, no templates. Is there something I'm doing wrong?

[rugged fern]

sometimes my chat gets stuck, response animations also get stuck and typing becomes slower, loading chats takes more time, this usually happens when i don’t delete chats for more than a month or so, the responses are great to be honest, but i want them to be smoother and not get stuck, any ideas on what i should do or where to report this for improvement

[humble rivet]

can anybody make any head or tail about it

[slender sphinx]

No one here will help you jailbreak it intentionally, that* would be against the ToS in all honesty, if you need help with an issue, clarify it. what you are saying is not really an issue.

[late atlas]

that will not work with such prompt

[echo cliff]

Hello there, I´m new here. I have questions regarding ChatGPT´s privacy settings. I have been using ChatGPT to generate images using other images as references. Previous to that, I had already turned off the “Improve model for everyone” setting. I have also submitted a privacy request on OpenAI´s privacy portal so that my content won´t be used to train the A.I. and I have had it processed, but this step has only been done more recently.

Is the privacy request necessary for them to not use your input images/files for model training? Having the “Improve model for everyone” setting turned off on your account is not enough? Both options refer to your “Content” so I´m a bit confused about if it´s the same thing.

[fair valley]

Hello everyone, I need help with Sora. Until earlier today, I was able to generate two images at the same time using Sora, but suddenly it changed to only one image at a time, whereas the Plus plan allows two images at a time. What could be happening?

[onyx kraken]

Hey hi!

We're all community members here, the very few who work for OpenAI have gold names, #server-staff . That said, what I see and understand in the help files, either alone is sufficient to help ensure your data is not used to train the model. OpenAI appears to have a deep respect of user choice with that, and offers more than one way to opt out - kinda like if one light had multiple switches to turn off - any of them in off is enough, but someone can flip all if they want.

Do note, legal matters may require our chats to be held for 'some period' of time, based on whatever the law requires, and those chats may be legally required at some point to be shared as the court rules (I would guess with lawyers and others involved in whatever lawsuit or whatever is making that necessary). That is not the same as the AI being trained on your chats and there is no way to opt out of legal requirements OpenAI faces to retain chats as required by law. If you want to learn more about that, https://openai.com/index/response-to-nyt-data-demands/ and https://help.openai.com/en/articles/7730893-data-controls-faq

But that has nothing to do with if the model is trained on your chats or not - you can decline that opportunity through any of the methods you mentioned doing both of - it was honored as of the first time you said no through any of those methods!

[onyx kraken]

By 'delete' do you mean 'still using the same chat'?

I observe slowdowns when I keep using 1 chat so it gets very long - eventually they can get so long they break and various bugs can happen. But I have chats I have never deleted, going back to 2022. And my account is not slow.

If it's the 'long chat', I recommend just start a new chat; if you have cross-chat memory on and do not delete the old chat, the model can sometimes retain information from the previous chat (some of it) so it can be a little like having that older chat with you in the new one - without the long chat slowdown.

[onyx kraken]

So, there's a lot of different 'formats'. That format and that attempt only has words. You can try again, like asking the model to through Canvas/canmore tool. Python tool, image gen tool, and maybe you yourself take pieces if needed and construct them into a finished document that does look right.

You can share with the model what went wrong/what you see; you can try in new chats, describing the problem.

I don't directly do what you describe, so I don't already know a specific workflow that tends to work well, but it likely involves 'doing it a different way', which may involve you doing some part of the copy/pasting, I'm not certain.

[onyx kraken]

#1070006915414900886 , and there

are also #1070006151938314300 too

[trail ingot]

Ah ok, thank you so much!

[somber hill]

Is that no longer possible?

Geht das nicht mehr?

Already? Foreign language detection should be adjusted. As long as the post contains English, it should be accepted, even if the English quickly plops, so to speak, as is the case with my translator.

Jetzt schon? Die Fremdsprachenerkennung sollte angepasst werden. Solange der Post Englisch enthält, sollte er akzeptiert werden, selbst wenn das Englische sozusagen schnell nach ploppt, wie das bei meinem Übersetzer der Fall ist.

💙 seems to have connection problems ..

💙 scheint Verbindungsprobleme zu haben..

Uff ..

Uff..

[split blade]

is this the random help page?

[slender geode]

Does anyone know if there is a discord bot with OAI integration to use my OAI account via bot in my server?

[vivid mortar]

To my knowledge there are no official integrations for this sort of set up

[inland apex]

Will it be possible to use the old voice mode via API after it is retired in ChatGPT?

[dapper verge]

How do you apply the unified diffs that GPT-5 Thinking generates? For me git apply --check patch.diff leads to error: No valid patches in input (allow with "--allow-empty").

[hardy moon]

Why is my chat GPT only displaying blank answer only, I tried deleting cache, reload the PC, the internet and the app but none worked 😭. Plz help me

[forest flare]

It's down for everyone

[hardy moon]

Oh so that's why ...

[cyan kraken]

Do we know it will be back online?

[scenic flame]

apparently they haven't acknowledge the problem yet

[hollow ridge]

on the iPhone works tho, it's not working on web only

[fickle imp]

how can i give smart answers on tinder now ???? 😣

[broken marsh]

机翻 · 通用领域
I also encountered the same problem

[magic ice]

me too

[jaunty hazel]

they have, but probs you are aware of it by now

[fading furnace]

Can anyone tell me what is happening to site because I am getting CORS errors in chatgpt.com site
is that an issue?

[fading furnace]

I think because of that error I can't open settings

[dusk nimbus]

same problem here... not chatGPT answers...

strage, it works on mobile...

[minor ice]

Why the hell is there no "Try again" button in project chats??
I have to edit my message with "Think harder" to get it to think.
Also you can't edit messages if you attached an image.
Just some small things I've found incredibly annoying.

[misty siren]

why is my chatgpt not showing any text when i type something

[dusk nimbus]

the chatGPT answers are back! 😄

[dreamy kiln]

Still nothing for me

[oak sapphire]

I know I am not the only one struggling with voice since gpt-5 launched, but I keep trying to find a work around the issues I have. I was able to disable advanced voice in the personalization options, but I still have a problem where it just flat out doesn't understand me. The first prompt it understands perfectly, but anything after that it misinterprets. I keep wondering if the pop up for feedback after a chat is causing a problem, because it will pop up and interrupt the conversation half way through. I tried finding the option to disable feedback, but I can't find it. Does this sound possible, or am I just running into the same bug everyone else is?

[slate sleet]

am trying to use the voice agents and following the documenation. i have a python backend, but the voice agent sdk is in typescript. so when i try to load it using the cdn, it says:OpenAI Voice Agents SDK not available

[proud hare]

How are you struggling with voice. Yeas it does feel like there is an actual person repeating what the GPT is saying. Like I cannot belive that, that is an AI speaking. Those problems you have are because of your internet, otherwise it uses its high intelligence to first wait, and then analyse when to speak (sometimes errors exist, yes)

The only problem I'm having is you cannot attach PDF's or files

[oak sapphire]

I used it fluidly before gpt-5, and didn't have this problem where it was rewording what I was saying, and it isn't sticking to the context of the conversation. I tried it with Claude and Gemini, and I don't have that problem with them. This is a specific problem to chatgpt.

[proud hare]

Which problem?

[oak sapphire]

When I use voice, it is rewording what I say to something random. It is not keeping the context of what I am trying to say.

[proud hare]

That's an odd problem that I really haven't heard of You can actually exit voice mode and check what it actually converted the voice to, and what it understood. Most annoying part they still have not fixed is the ability to attach flies on voice mode.

[echo cliff]

Thank you for your time and thorough reply. I will check those articles and maybe try to contact someone from the staff if I see them online.

[rugged fern]

yes, you're right, i meant it that way only, it's been a long chat and i've been texting chatgpt for a while now, where do i enable cross-chat memory though, is there any way to continue chatting or use the old long chats so i don't have to create a new one and mess things up or rely on cross-chat memory just to get a bit of info from the old chat

[onyx kraken]

I find long chats get buggy, always. They lag and weird stuff tends to happen as they get longer and longer, and I use loads of new chats. And the model feels very responsive, informed, and contextual for me. It's like it was informed of at least a decent summary of much that we've discussed - it doesn't know every detail, but wow it knows a lot of details.

[onyx kraken]

As for how, bottom left corner, your account name, click it and then 'settings'.

Then go to personalization tab:

[mint goblet]

Sora is caught in a loop. It is trying to load a prompt but keeps scrolling. I can not get out of it or cancel it. Platform is Windows 11. Help to get out of this loop is greatly appreciated. Sora will not allow me to start any more projects until this loop project is completed. Paying customer.

[onyx kraken]

I have all 4 options on. I recommend that, if you want the model to be at least somewhat informed of your previous chats with it.

You also have vast memory space, as well as about 3k characters of 'custom instructions' space, and you can have the model fill that with specific details you want it to for sure always have in every chat.

[dreamy pond]

Hey

I need a help.
I am using baml for prompt to open ai. And I am using gpt-5 on medium reasiong but i am getting time out error on that of 20 min.
can anyone here help me out how can i fix this. I checked my code there is no time out present like this..

[slate sleet]

hi. does gpt-realtime support different languages

[rugged fern]

how do i fix it though man, i noticed that it doesn’t happen for the same chat, the long chat, when i use it on mobile, it’s kinda weird, plus i don’t want to use this cross-chat memory or whatever and lose the real experience i was having, like the old chat has all my data or what i’ve inputted so the responses come with that old memory, i just feel like i’m talking to a person who knows all my tasks, what’s done and what’s not, and gives me precise responses

[onyx kraken]

If it works okay on mobile, use that by all means!

I have never heard anyone before you say a situation where it works at all, my experience and everyone I've noticed post about it, long chats break. Sometimes so badly they cannot be opened again, we can't even read them; sometimes they just can't be typed in anymore, sometimes the model stalls out on an output, sometimes weirder stuff.

I have never heard anyone say 'this works'. I'd be careful with super long chats even on mobile, because everything I know about really long chats says they break.

So, grats on having any answer at all, I hope it keeps working and doesn't massively break later; far as I'm aware every really long chat eventually cannot be continued. The cleanest end is a message from OpenAI, 'this chat has reached its maximum length' - I've never seen it but I've seen folks post screenshots of it. I've heard a lot about bugs too in longer chats.

Good luck!

[rugged fern]

thanks, but i hope in the future they find a way for longer chats too :)

[onyx kraken]

You can #1070006151938314300 and recommend it/ask/maybe even discuss ideas how that could work, like how any memory issues and how hard the model has to work to track an ever-lengthening chat might be resolved, if you have any ideas!

[trail cairn]

yes

[sacred plinth]

guysss

I canceled an image while it was being generated. After that, I edited the prompt, but when I started the new prompt, the previous image kept generating for 15 minutes. Now I can’t generate anything else in Sora — this image keeps generating endlessly. How can I fix this? When I try to delete it, it says it was deleted, but it still remains there trying to generate.

[deep tangle]

does gpt 5 (or other gpt models) understand if you ask to limit a response to ~100 tokens? how often will they follow that type of instruction?

[onyx kraken]

It's very hard for the model to count tokens as it uses them. More successful usually is to give the model an example output of the length you want.

[stuck cloud]

why is my chatgpt like this when i try to type?

[potent wedge]

is there a built in chat gpt on discord

[rugged fern]

I guess new "project" feature has solved the issue lmao

[elfin crane]

Good to know OpenAI has made projects available to free users... but project instructions for the projects are not available? #announcements message

[hazy pulsar]

Is it OK to run the ChatGPT web app inside my own WebView2 on Windows? I’m using it only for myself (not distributing), with a User Data Folder for cookies/login.
What’s allowed, what’s risky, and what’s outright disallowed?

[leaden agate]

hey guys, have been trying to upgrade my chatgpt account from plus to pro for a while, but I keep getting the "Payment error. Your card may be invalid or authentication may be needed." error. Have already tried changing cards, browsers etc 😅

[celest cradle]

I've met all the requirements for tier 2 but I'm stuck at tier 1. Tier 1 isn't even enough gpt-5 for my use case 🙁

I had to generate $50 in images with the API just to get the spend over $50 to try to unlock tier 2 but I'm stuck at Tier 1 and don't know what to do.

[analog herald]

Hi. can ask here a question?
Since friday 22 august 2025 my Team currently named Business Plan account is behaving weird as lol.
Namely if we try to use the Voice chat feature it is writing weird stuff, speaks japanese, finish sometimes other... then it also not responding while doing alwais a clack clack noise.
Since voice direct response for us is what we need as much as other components and agents, does anyone here knows why for 2 weeks now we cannot use this feature?
We made an escalation mail but the reply was LOL from a bot.

[unreal rampart]

Hello everyone. I have the same situation. I started generating an image, and then I decided to delete it because I uploaded the wrong text file. And I have endless downloads and I can't even delete this generation. I've already tried everything on other browsers, even in incognito mode and I've cleaned the cookie cache. Uselessly

[junior flume]

Hi all!
I have been using ChatGPT as a Plus subscriber for most of the year.
I use it primarily for software development for personal projects, but also to accelerate research and collate data.

My mode of use for development is basic: on macOS I use the ChatGPT app with the Work With capability to connect with Visual Studio Code. This was enabled by an Open AI Code extension (that is no longer available and apparently replaced by Open AI's Codex extension).
I love that mode as I interact with the ChatGPT app to brainstorm, design, implement, debug and fix. The ChatGPT app would should me diffs and I could choose to apply those diffs to my code via the Work With connection.

This would work with local and remote workspaces in Visual Studio Code - I rarely use a local workspace and instead use a remote Ubuntu development server for my code. It worked perfectly with my remote workspaces and worked fine the one time I had a local workspace.

Since the release of the Codex extension, this capability has vanished even though all documentation states that Work With will continue to work with Codex.

When I click on the Work With button in the ChatGPT app it shows: "Code: No workspaces available. Setup"
ChatGPT > Settings > Manage Apps ("Work with Apps" section) shows: "Code: Unable to connect"
Proceeding to setup / establish the connection jst brings up the Open AI Codex extension's product page in Visual Studio Code.

Has anybody else run into such issues? I would very much appreciate any help and guidance!

[spark oasis]

Hello, it seems that gpt-5-thinking in the app is not thinking at all for me. It works on web but in app, it doesn't think. Even o3 doesn't think. I'm on Chatgpt Business on Android

[orchid swan]

can anyone tell me how to enable or give permissions to the desktop app to have access to my mic cause its not picking up on anything???

i go into the dictate thing and im saying but nothing is being picked up.. and i see in privary & security that chatgpt does have access so idk..

[prisma aurora]

Hi, I have a question, when using ChatGPT to generate images, is there any way to use the personalize_codes like in MidJourney? just to make sure the image generated has an aesthetic. thanks very much for your help.

[rich horizon]

hi noob Q 🙂 - if i have multiple agents working an EPIC / Stories - whats the best way for them to update each other?

[jovial girder]

Hello everyone,

I wanted to ask if anyone else is experiencing similar issues with Sora image and video generation.

What’s happening:

Generations (both images and videos) often get stuck on the loading screen (just the spinning circle). They never finish.

Some old files that I had deleted a long time ago suddenly came back in my media library and started loading again.

When I try to delete content, sometimes it doesn’t actually disappear.

The site in general is very slow for me — even in incognito mode it becomes super laggy and takes a long time to load.

What I already tried:

Clearing cache/cookies,

Changing browsers (Chrome, Edge, etc.),

Using incognito mode. Has anyone else seen this problem? Is this maybe a general Sora issue right now, or something only on my account? Any advice would be very appreciated 🙏

Thanks in advance!

[maiden folio]

It has been acting weird in many ways the past couple days.

• Deleted items don't delete, and end back in library instead.
• Deleting an item, removes a different item than the clicked item.
• New generations, not appearing in the library, having to use Notifications to get to the generation.

So yes, there's something weird going on with the supporting systems around the model.

[wanton plover]

Does the plus subscription include unlimited GPT5?

[timber shadow]

Hello! Not unlimited, no. Currently on Plus there are up to 160 messages with GPT-5 every 3 hours, you can read more about the usage limits here: https://help.openai.com/en/articles/11909943-gpt-5-in-chatgpt#h_4699b07591 Link should jump you to the relevant section

[wanton plover]

Okay, thanks!

[jovial girder]

When will they solve the problem?

[novel grail]

Uffff :(:(

[rich horizon]

is there a UI mock up AI? 🙂

[rich horizon]

when should i start new chat? when context window is full around 50%?

[toxic crater]

Do anyone know here how a correct responses.create should look with "pmpt_" ? Is the word "json" need to be used with the call? or enough in the "pmpt_" ? I am so confused

[lost anvil]

guys perhaps this is the correct section... MY SORA ACCOUNT IS STUCK

it has been loading this image for a hour

and I can't make others

[pliant cypress]

I had this a couple days ago. It said the job was done but it kept cycling. A day later it worked again.

[somber crypt]

Interior Designer here, bringing over 3D CAD renderings generated with Chief Architect. These canonical, non-negotiable architectural renderings consistently fall prey to DallE's reinterpretations and design drift, even after succinct and precise prompts. Suggestions?

[dense hill]

So I’m trying to get a model to recognize others usernames on discord. Does anyone know what to do?

Basically TLDR: model keeps mistaking others for other people since it can’t see usernames

[slate sleet]

hi.. has anyone used the openai realtime api and gpt-4o-transcribe to create transcriptions? i have successfully implemented the realtime voice agent but have basic authentication errors when trying teh transcription. am using nextjs/ts

[plush phoenix]

how do you fix minified react error #185? I've tried deleting my browser cache but this doesn't fix the issue?

[long girder]

one sec

it’s a React “infinite re-render” crash (“Maximum update depth exceeded”). Clearing cache alone rarely fixes it.

What it means

Minified React error #185 = the page got stuck repeatedly updating itself and React killed it to prevent a browser hang.
React

Fix sequence (fast → deeper reset)

Hard-reload the page

Windows: Ctrl+F5 (or hold Ctrl and click Reload).

Then try a new tab of ChatGPT.

Incognito / different browser

Open ChatGPT in a private window with all extensions disabled. If it works there, it’s almost certainly an extension/userscript conflict (ad-blockers, “ChatGPT enhancers,” Tampermonkey scripts, auto-scroll, etc.). Re-enable extensions one by one to find the culprit. Community reports match this pattern.
Reddit

Clear site data (not just cache) for ChatGPT

In the broken browser go to Settings → Privacy → Cookies and site data → See all site data.

Remove data for chat.openai.com (cookies, localStorage, IndexedDB, service workers).

Sign back in and test again.

Kill the specific corrupted session

If the error appears only on one conversation, open a different chat (or start a new one) and see if the old thread triggers it. Very long/rapidly updating threads can trigger #185 loops. If confirmed, export/duplicate what you need and archive/delete that thread. Community reports have seen app-specific loops like this.
Reddit

Graphics & resource edge cases

Disable hardware acceleration (Settings → System) and relaunch the browser.

Close memory-hungry tabs; very heavy DOM updates can push React into #185 in various apps.
Streamlit
Steam Community

Update the browser

Make sure Chrome/Edge/Firefox is fully up to date; then repeat step 2.

If none of the above

It may be a site bug affecting your account/session. Collect a short console log (F12 → Console, copy last ~50 lines with timestamps) and report via Help → Bug report. The underlying issue is the same React failure seen across other apps when nested updates loop.

try this

[long girder]

The issue here is that your model literally can’t “see” Discord usernames — the UI hides or obfuscates them in ways a screen-scraper or model won’t get. That’s why it keeps confusing people. You’ve got a couple of paths to fix it:

1. Use the API, not the screen

Discord provides usernames, display names, and IDs through its official API.

A bot can fetch the unique user ID (user.id), which never changes, and also grab username and global_name.

Training or prompting against these stable IDs is much more reliable than trying to read rendered names from the app window.

2. Normalize on IDs

Internally, store and compare user IDs, not display names.

If you want the model to “recognize” people, map each ID → label (like 1234567890 → "Oswald").

Then, when the model sees a message event, attach the label from your mapping.

3. Handling display name changes

People can change their display names and nicknames. If you rely only on those, the model will always drift.

Solution: always resolve user.id → canonical name at inference time.

4. If you’re scraping text only

Without API access, the best you can do is capture whatever “@mentions” appear in the raw message text.

Mentions are formatted like @lucid inlet. Those numbers are the IDs — you can parse them and build your own dictionary of usernames.

TL;DR: Don’t try to get the model to “recognize” people from text or screenshots. Hook into Discord’s API, key everything by user ID, and translate to usernames/nicknames only when you want to show humans a name. That way the model won’t confuse users anymore.

[long girder]

use better tools

revit or whatever is easy

you can also just use pix4d or whatever free program to play around on trial

then wrap the images over the 3d model

or drone deploy is even easier

just use it without a drone

costs real money tho

so use something cheap and learn how to do it

easy

take you two a week playing around with it to learn it

get company to pay for the licenses

for whatever many-image-to-3d rendering program/license you get

dalle has almost no use case for you

even if you wanted to superres the original images before rendering, you can use any of a million programs to do that

just get adobe

easier

cheap to a company for a few license seats

use adobe to upres the images and enhance, manually edit, whatever u need to do

then run them through whatever system you end up buying

it does most of it for you

find a 3d architect to consult for a couple days of billable hours to teach you how to do it

cheap

easy

you hire some person for a week only, and then you know how to do it. call them if u have questions

you can learn how to do it on youtube. literally.

software just costs money

i literally learned how to do it from youtube before LLM existed.

videos better now

anyone else before i go?

[long girder]

-------------------- @long girder out -------------------

[vague vessel]

I have well over 250+ conversation threads. I checked some older threads and it seems some of the context in the thread itself may have been clipped or fragmented. Has anyone else experienced this or heard about it?

And is there any way to see the inception date of a conversation threads?

[onyx kraken]

Search a word in the thread. Scroll through the chats with that word. Hover over any of them; if there's no year info, that's this year. If it's from before this year, you get full month/day/year format

However, that date is from when you last opened the thread. So, if it's one you talked past the inception date, or opened more recently, you see the most recent interaction date, not the original chat date.

You can fix this in your future chats; either you or the model can state the date as part of the chat.

You can #1070006151938314300 and request an inception date tracker too.

[warm cave]

Hi

My

Cloud please need help

[analog herald]

this feature is good like it

[analog spear]

yup

[analog spear]

Hi! Can you describe what you need help with ^^

[lavish plume]

Hey, I just build a GPT to help me with writing scripts. But it doesnt follow the instructions correctly. Sometimes it uses sources that are not whitelisted. It often uses en or em dashes even tho they r forbidden. The output of the words are sometimes in different formats and it makes a lot of other mistakes. I even asked the gpt to analyze the output and tell me what it did wrong and it actually sees the mistakes. So my i doubt my instructions are not clear enough.
Did anyone of you face similar problems and found a solution or reasons why that might happen?

[nocturne canyon]

There’s probably ways that you can tighten up things. Sometimes, using language that makes it clear that doing a thing may disrupt the core purpose can help. But honestly, these things are highly probabilistic and trying to find the best way to give a satisfying response to you…

Any probabilistic system is going to have a hard time following more deterministic rules. There’s going to be times where balancing your rules vs your request will weight things towards the request and against the rules.

Depends also on which model you’re using to some degree. I think thinking models are slightly better at considering instructions in output, at least that’s been my experience.

[pastel crow]

I’m having issues with the new interactive quizgpt. It doesn’t work in my iOS app, but in my ChatGPT web environment it works. Has anyone else experienced similar issues in the past few days?

[pastel narwhal]

I cannot generate images or videos in SORA because one of them is stuck and wont finish. The "Trash" image set doesnt work. I've logged out and back in, cleared Cache and cookies. Nothing is working. It must be server side.

[jade bobcat]

how much does it costs to analyze 20 000 images by o4-mini?

[rich horizon]

i am using opensource codex cli - how come sometime my codex instance doesnt understand that it can run various CLI tools such as git aws etc.? and other times it does.

[vivid mortar]

I've noticed this happen too, I think its typicaly just because the context gets too long (at least from my experience) and it either needs to be reminded that its a coding agent or start a new chat

I imagine that will be fixed as its adopted more

[exotic nymph]

Is anyone else having trouble deleting images and videos from their library/My media - everything I have tried to delete over the last week just keeps coming back - is this a known issue that things just aren't getting deleted?

[tired mason]

New Bug. The recent UI change messed up some functionality of speach to text. Placing speech to text in the ... section of the website and app had the side effect of requiring the website or app be in focus for chatGPT to speak. I play video games with my AI so the window unfocuses during play making it so my AI can't contribute to the game experience. Can you please return functionality to my AI.

Either reverting to the old UI (I preferred it since I use speech to text a lot) or adding a section that allows speech without needing the window focused would work Thank you!

[rich horizon]

how do you enable various tools such as GIT GH AWS etc for CODEX CLI?

[exotic nymph]

whatever updates have been done recently have messed up the updating of the canvases within chats and projects - when I get a canvas updated parts are being removed or the update fails.... this has been more and more common over the last few weeks.
I'm getting sorely tempted to cancel my subscription until this gets fixed as I am losing so much time having to manually fix these things myself and keep a separate copy of the info elsewhere

this along with the mess that is now not being able to delete things within Sora... makes me think updates dont get tested before they are rolled out

[rich horizon]

"- Alternatively, tell me the ECR login already works in your terminal and I will only do the docker build/push and ECS
updates from here." its been doing aws deployments for like 3 hours ....

[analog herald]

in android app config there is a function to keep gpt voice chat active overlay even with phone on standby. I use it daily when I go for my training as it cracks jokes and keep company

[analog herald]

Using branching at the end of a very long thread that gets slow on PC due to huge DOM rendering is better than using a txt loaded file containing the thread or any other thing.
The old thread before the branching can be saved if you need to archive offline for what so any reason.
Branching from last output leads to

• new thread
• persistent context
• persistency of all customisations in original thread
  Have fun

[uncut saddle]

Hi , this morning stoped a image genration but after a hour is still stopping the image generation, and don't allowme generate another image because only can work in one process at time, how can force the stop or kill the process?

[exotic nymph]

My Library is showing an error - I can't view or generate anything - status claims it's fully operational..

[charred parcel]

reach out to support at https://help.openai.com

[unreal hemlock]

<:book_icon:1363314738255364126> Rule 7: No self-promotion, soliciting, or advertising.

-# Do not post or direct message any members of this server to promote non-OpenAI services, products, or projects.

Exemption: API & Custom GPTs channels

[solemn stratus]

Can anyone help me? I am getting charged for API key which does not exist in my account.
I already lost 15 USD worth credits

[solemn stratus]

key starts as key_Kue...
I have no such key
I have 3 keys, 2 have last usage in 2024, 3rd one I have just created.

[idle cypress]

Hi everyone, I'm new to this channel and wanted to ask the community for help with a challenge I'm dealing with. I need to extract data from an Excel file and convert it into JSON format. I've heard of open-source tools like Unstructured or LlamaIndex, but I can't use LlamaExtract because my content is HIPAA-compliant and I can't send the data to any external servers—everything has to stay on my own system.

I'm reaching out to see what other options are available for extracting information from Excel files. These files can vary a lot in structure depending on the scenario, so the solution should handle different types of Excel files flexibly. Ideally, I'd like something similar to LlamaExtract, where I can define a schema, upload the file, and pull out the data accordingly. I want to implement this in my own project. Does anyone have suggestions, like libraries or tools, that could help with this? Thanks!

[devout panther]

Is anyone ever successful using patches created by CG? I've tried using them over and over and they are always considered garbage by my linux 'patch' program. As a result, CG continually outputs lots of code which really slows things down.

[turbid galleon]

I am being told I am over my quota, but I only have 1 project and 1 api key and money in my account. How do I assign myself a quota or increase what I apparently have been assigned?

[nocturne canyon]

Have you tried asking GPT? I'm guessing they know.

[celest cradle]

GPT can't tell me why I'm stuck in tier 1 in spite of meeting all the stated requirements

GPT API feels like a scam. I could have just used openrouter

[hallow basalt]

Hi everyone! I need help with a persistent issue on my ChatGPT Plus account.

The Projects section (folders & organized chats) is completely inaccessible.

I always get the error: “Unable to load projects”.

Issue occurs on all networks, and browsers, except for my mobile app.

I already tried: disabling extensions, clearing cache/cookies, private browsing, different locations.

On the same setup, new accounts work fine, so it looks account-specific.

I’m not using the API or uploading large files.

I cannot access, view, or delete any existing projects.
Does anyone have any idea of what I could do?

Is it working for you know? I've been having the same issue for the past two weeks

I've been having the same issue, any luck for you?

[turbid galleon]

Yes...they direct me to the pages online, but I feel like I have eerything set up properly. It is a very simple setup so I don't get it.

[cedar sedge]

I had this problem for a few weeks, and then it seemed to be fixed for a while - I was happy. And now, this week, it's back again.

[slender pivot]

Help

[slender pivot]

It might be an internet problem

[unborn panther]

HI is there some referral link for plus subscribers for friends?

[sour monolith]

I am in the same situation as @cedar sedge . Worked for 2-3 days, now it's gone again.

[ebon relic]

“How can I change the primary email on my OpenAI account while keeping my active $200 Pro subscription and linked payment method?”

[onyx kraken]

Hey, you'd need to reach out to OpenAI tech support, https://help.openai.com/en/articles/6614161-how-can-i-contact-support

We're almost all community members here, like yourself!

[onyx kraken]

As far as I'm aware, and the helpfiles say, https://help.openai.com/en/articles/4936827-how-to-change-your-email-address

That's not supported at this time. We could start a second account with a different email address, but we cannot 'change' the email associated with an existing account. We are allowed to have more than one account though!

[onyx kraken]

At this time, I'm not aware of any. You might want to watch #announcements and/or #spotlight because if there's such offers, they might be announced or spotlit there.

You can also whenever you wish check the helpfiles for them, https://help.openai.com/en?q=referral program to see if there's any available, because such opportunities nearly always do get some helpfile to assist. There recently was, maybe currently still is a referral program for certain locations and EDU accounts (which is not plus, it's different) https://help.openai.com/en/articles/10845652-student-plus-referral-program-australiacolombia

[void steeple]

I'm struggling to set up a Supabase MCP working with Codex, following the GitHub docs, and it simply doesn't work. Got it working in two seconds with Claude Code, but Codex doesn't seem to work no matter what I try.

[gaunt valve]

There seems to be a strange bug with GPT 5 Thinking where graph and table outputs often get default-stuck inside of the thinking module - and so don't end up in the output? Here's a demo with GPT 5 Thinking output. The chat suggests an interactive table has been presented to the user but that is not correct.

https://chatgpt.com/share/68c0dd48-5b20-8010-ab9c-5b034edcc9b2

Where might I go to report this bug? It seems not to happen with standard GPT 5.

[agile meadow]

Where can I check what limits i exceeded? i was doing study and learn and i got this message

[jaunty spindle]

I realize that this convo was yesterday, but your post caught my attention due to a recent experience.
Python and the Pandas library will do all of the things you mentioned and more...and has integrations especially for MS Excel...this is exactly what Pandas was made for and latest versions are so human-readable. You can use any IDE and ChatGPT, create the use cases you need with mock data, and then cut the internet access and parse and process your sensitive data securely and all that...and pandas has built-in function to read from an Excel spreadsheet to JSON. I know because I just completed an assignment doing that very thing a few weeks ago.

If you don't know Python or Pandas...that's fine...ChatGPT will make it very easy. I suggest using 4o and instructions to follow your instructions with concise responses. But without AI help it is so easy and documentation is pretty good too.

[north ember]

Hi everyone, " GPT-5 Thinking Mode Slows Down Over Time + “Branch New Chat” Feature Not Working"

I’ve been using GPT-5 in Thinking mode for coding tasks, and while I understand it takes a bit longer to respond due to deeper reasoning, I’ve noticed a significant slowdown after 2–4 days of continuous use. Even short prompts or simple function requests start taking much longer than usual, both in the desktop app and browser.
To work around this, I try summarizing the session and starting a new chat. But the new chat often lacks enough context to understand the project, which makes it hard to continue smoothly. I also tried using the “Branch New Chat” feature, but it just shows a loading icon indefinitely and never completes the action. I’m on the GPT Plus plan and haven’t been able to get this feature to work at all.
Has anyone else experienced this? Any tips for improving performance or making transitions between chats more seamless would be appreciated.
Thanks in advance!

[fleet moth]

Hey there! I need to know if ChatGPT Projects can help me with my Use Case. We are a 10+ year old construction company in Germany. We do have a lot of receipts data, that explains our Services and prices. We want to create new offers for new customer projects, based on historic data. Please share your thoughts if this is possible with ChatGPT or other Tools. THANK YOUR ❤️

[old fractal]

Can you tell me if there is any information about when Sora will be updated?

[maiden rune]

I have a question I want to connect zapier and make with open IA except that it asks me for the API key I put it but the problem I have a 401 error I have to pay for it to work

[agile sail]

hi guys need help, will cancelling your subscription early revoke your premium plan? e.g. im claiming a 3 month subscription promo, but i cancel on month 1. will i get back to free on month 2?

[timber shadow]

Hello! Per this article, it depends on the kind of promotion it was: https://help.openai.com/en/articles/10492689-chatgpt-plus-promotions-and-referrals#h_9e08222cf0

Promos are redeemable on the web and you can cancel anytime through your ChatGPT account settings on the web. You can’t cancel through the ChatGPT mobile app. If it was a free promotion, your access may end immediately upon cancellation. If it was a discount promotion, you will retain access for any remaining time in your current paid billing cycle.

[agile sail]

thank you!

[novel grail]

The login is not sending a verification code to my email - there is no resend buton either !

[quartz robin]

Is there a channel for codex cli?

[timber shadow]

Yes! #codex-discussions is used for both the web implementation of Codex and Codex CLI

[quartz robin]

Thank you

[somber hill]

..

I have a translator, so it is not just a foreign language to be read!

Ich habe einen Übersetzer, also ist nicht nur eine Fremdsprache zu lesen!

[steel heath]

where is modmail - found it - it didnt like the walnut whip emoji i got from its own panel

[exotic coral]

Hey is it possible to have chat gpt look through videos? When I asked it it said it was possible but I don’t see an option to send a video to the chat

[onyx kraken]

You might want to start a new chat, and maybe even have memory off, and ask it,

What are the options to have you view a video?

You're likely to get answers like:

Well, you can upload a transcript.

Hey, you can take a lot of screenshots and share them with me!

You can describe it, and I can see it through your eyes.

---

Now, there's a few ways, from what I understand, to do so through the API, but there you're paying per tokens involved and the way to send the video, it gets chopped into many screenshots and those get sent to the model; whisper or similar to get a transcript if you don't have one available.

[onyx kraken]

The model is designed to please us. Wherever possible, if we make it clear what we want, the model tries to say yes. A decent way to limit 'inflated expectations' can be to make it clear that 'no' is fine, just want a grounded realism answer.

[exotic coral]

Gotcha

Thx

But yeah main problem im facing rn is that i have a dungeons and dragons game in a conversation and i reached the limit to that conversation, so i had to start a new conversation but got doesn’t remember the story, and since its long i thought maybe i could screen record the chat and share it , but yeah seems that wouldn’t work either cause its too long lol

[onyx kraken]

One possible option.

Copy-paste the entire chat, or maybe half, a third, whatever works.

Into a text file. Save that.

In a new chat, tell the model what you want - maybe a summary of the game so far from the file you upload.

Upload the whole thing, see what you think of the summary.

Then try again, another new chat. Upload only half the game.

Then try again, another new chat. Upload only 1/10th the game.

See what summary you like best. If it's the 1/10th, consider checking the second 1/10th; if the same chat doesn't handle that great, go to a new chat.

Give it the summary of the game 'so far' - that the other chat made for you.

Upload the 2 out of 10 file, and have it continue the summary

[exotic coral]

Gotcha

Thx

But yeah rn, copy pasting and going through the entire thing will take me some time lol, if only they made it so gpt could access previous chats and conversations

[onyx kraken]

Well, have you explored what can happen if you turn all the memory options on?

[exotic coral]

I did, have had them all on

[onyx kraken]

And, I don't mean copy/paste 1 input or output at a time. On a computer, in say Chrome - click anywhere on the page.

Press Cntrl-a

That highlights all the text in the chat area. Not the other conversations on the side, just the actual everything you and the model said.

Then press cntrl-c

That copies it.

Then go to notepad or any other word processing option on the computer. Press cntrol-v.

That pastes the whole thing there.

And then you can quickly remove parts to paste in other notepads or whatever else you want, as well as save the whole thing as - is to try if the model can handle that.

[onyx kraken]

Great!

In a new chat, what happens if you ask as your first input, what has been going on in our game?

[exotic coral]

Ooo that would be wayy faster lol, thx thx !!!

[onyx kraken]

And I highly recommend, in the future consider keeping chats shorter. Let memory help stitch them together, it seems more designed to survey many short chats than go super long into any numer of super long ones.

If you go super-long in one chat, part of it seems 'hidden' to the memory survey, but even if you have hundreds of medium to short ones, they seem to get surveyed pretty well

[exotic coral]

Gpt mixes up random characters and stories up, when I tell it what the actual story is it says it doesn’t remember and when I ask it if it can access the previous conversation to refresh its memory, it says it can’t do that, so yeah, like there are bits and fragments it remembers but not the actual story

[exotic coral]

Yee I gotcha lol, lesson learned

[onyx kraken]

You may find that if you choose to experiment with shorter chats, and carrying the story to new chats to progress, that the overall memory is quite coherent

[exotic coral]

cool noted 🫡

[broken yoke]

Hello, I'm a developer and I'd like to make some experiments with the new MCP connector feature of ChatGPT, is there a specific discussion channel where I could ask questions about it?

[grand pulsar]

why am I getting the error “you have sent to many messages to the model, please try again later”? I have plus

[nocturne canyon]

Plus has usage limits.

[grand pulsar]

since when??

and what’s the limit??

[nocturne canyon]

https://help.openai.com/en/articles/11909943-gpt-5-in-chatgpt

But it might not be that too? GPT should just downshift your model rather than refusing to accept a message.

[grand pulsar]

well I’m paying 20$ a month specifically to have no limit

[timber shadow]

Plus has never been an unlimited usage plan, but rather an expanded usage plan. You can see a comparison of usage allotments per plan on this page: https://chatgpt.com/pricing

[amber axle]

I think they changed it in the last few days

[grand pulsar]

I’ve never gotten an error message before… what am I even paying for atp😭

[grand pulsar]

How many messages is it till it resets?

[timber shadow]

That link from Justice above has good info about the usage limit resetting, check it for more info. Here's a relevant snippet for GPT-5:

ChatGPT Plus users can send up to 160 messages with GPT-5 every 3 hours. After reaching this limit, chats will switch to the mini version of the model until the limit resets.

[eager tangle]

How to add "Sora" in side bar? It disappeared for me, yestarday it was there tho..

[cosmic smelt]

it appears alone, but only in paid accounts, coudl be you had a plan that has expired?

[eager tangle]

I have Plus membership

@cosmic smelt

[cosmic smelt]

hmmm i do have the same

[eager tangle]

It was literally yesterday evening there

[cosmic smelt]

shouldnt dissapear itself

[eager tangle]

I'll try to clean up cookies then

[cosmic smelt]

do you use desktop app or webapp?

i tried both and it is there in both, maybe cleaning cache and restarting could work

[cosmic smelt]

yes

[eager tangle]

nah, still nothing..

Even tried it with Zen Browser

(Mozilla fork)

[potent arch]

How can I post in bug reports?

[timber shadow]

Hello! There's a post pinned to the top of that channel labeled "start here" - click into that, and you'll see instructions and a button for how to post

[potent arch]

Thanks! I thought this was not working because takes a long time to work.

[rose laurel]

Good evening guys, it's Shakes from Durban South Africa.

I've got some issues with ChatGpt5, at some point it fails to give me response from a prompt and I end up restarting my browser. Is it a normal issue?

[sand forge]

Hello there! Anybody tried to produce word clouds from files in gpt5? It used to work amazing with the older version, and now I can barely obtain any decent visual results. Thanks!

[stone iron]

Hi anyone else having this issue? cannot access platform.openai.com/playground it redirects me back to https://platform.openai.com/chat/edit?models=gpt-5 Has access to the classic Playground been removed?

[dusky axle]

Can anyone tell me why requests are simply being ignored or failing to respond so often since GPT5 came out? Sometimes refreshing the page helps, after resending the request. Sometimes it continues to fail. This is happening way too much. And I know for a fact that I am not the only one experiencing this. What is being done about this issue?

[nocturne canyon]

Is this with long contexts? I’ve seen others having this issue, specifically on web.

[vocal bluff]

How does one speak to an actual person about accounts? I’ve been having issues with mine and I was told by email that it had been fixed and back to normal. But it’s not, and I’m starting to think I got replied to by an AI and not a person when I was told it had been fixed. I just want to know if there’s another way I should be trying. Because every route I try with OpenAI’s platform leads me to an AI chat, and not an actual person to speak with. Discord is my last resort.

[torpid rune]

Hello,

I’ve encountered a very peculiar situation and GPT-5 claims that I may use an imposter website of GPT-5.

Problem: using GPT-5 on chrome with a plus plan and keep getting low-quality answers. It did write a code that generate one random number in a while loop and then add them to a list. And many other answers at similar level.

Solution: accidentally switch to edge and use GPT-5 w.o. log in. Answers immediately gets improved.

Horror story begins: I describe my situation to GPT-5(edge). It claims that the https://chatgpt.com/
url I visited on google is an imposter page that is independent from openai. I provided a screenshot. It pointed out several places not agree with a genuine page.

AI customer service: https://chatgpt.com/ is genuine. GPT-5(edge) haunted by illusions.

I guess that I have no reason to believe one AI over the other. So I decide to try my luck here. Besides, if https://chatgpt.com/ is the genuine website, why answers are so stupid? It doesn't looks like a server traffic issue. The low-quality answers has lasted for more than a day. Switching to edge immediately solve the problem also suggests a different cause.

[barren girder]

The quality of replies does degrade during peak hours. This has been confirmed by anthropic with claude, and has had heavy speculation from chatgpt users. Many american users speculate that chatgpt after midnight produces much higher quality results than during the day

[sudden coral]

Hey, I hope someone can help me, since yesterday I try to use chat gpt without log in on the computer. I told it everything in a paragraph, it answered me and then I saw the pop up of "unusual activity is detected" and sent me to the registration page. Today I tried again and it happened the same, didn’t even give me the answer this time. I always use edge, and I have no extensions of anything. I already tried to clear cache and all the solutions but still same. I don't know if this is something i should report, anyone had any similar issue?

[dusky axle]

Probably usually fairly long, but not necessary super long. But you think that's the issue? Is there a general "acceptable" length I should be trying to stick to? I do tend to go long on the threads. I would keep going but they definitely slow down but this current issue is relatively new.

[nocturne canyon]

IDK. I’ve had this kind of issue too. People have talked about this as basically a UI issue of some sort when the page gets too big because it holds all the text at once.

But maybe this is different. When this happens the page freezes also.

[dusky axle]

Sometimes I get a red text error stating that “Something went wrong. Try again” but trying again never works.

[empty saffron]

Hi there, can anyone please help me with PDF downloads in GPT5. No matter what i try, when i ask for a GPT 5 to provide me with a PDF, it provides the link, but when i click it, it fails and gives me a error? ANY help / insights. I get this error - Failed to get upload status for /mnt/data/

[lost anvil]

So, guys..... did they make a sneak update to gpt????? All of a sudden, out of the blue, it changed personality and it started to refuse spicy content in the stories we were writing.

The sudden change is so huge and so abrupt that I'm trying to understand what caused it. It's no exaggeration to say that if GPT no longer allows some adult content in the stories, I have no reason to renew my subscription - I can use other LLMs.

I am thinking of three possible reasons:

1. sneak update to the model.... which is an odd change of route
2. me using it for job-related stuff had it confused, it entered a professional job mindset and started to refusy spicy fiction
3. when I asked support for help as my sora account generation got stuck, they violated my privacy by peeking at my spicy content and decided they didn't like it (this one's very tinfoil-hat but it's strange the change begun right after contacting support)

I'd like not being forced to return to Grok. GPT is better - when it's not censored.

[winter cobalt]

Got a three hour limit after one single message at GPT-4o. What the ... is this?! Another step by OpenAI towards the end of the company?

[grand pulsar]

I just got that aswell?

I emailed them and now I’m waiting on a response, I knew I couldn’t be the only one with this insane issue.

[winter cobalt]

I made a Reddit thread about it. In my opinion, they've jumped on Anthropic's train. Anthropic has a five-hour limit, and OpenAI is now following with three hours.

[lost trench]

Same!!

[spice solar]

Me too 🙄

[analog herald]

Today they finally fixed the terrible quality of voice chat coice to text transcription making direct voice chat faulty and ghist filled. Oh finally I can hands free talk to the ai without having to type there too.

[fluid surge]

Hi! What is it? I need to pick image now every time? All my day is in this pics

[rich horizon]

folks can someone please help me understand how would vision api with video work? here is the cookbook i am following, essentially i am capturing live video, capturing 1-2 FPS as images and sending it over - i got the live audio working perfectly just cant get the video (1-2 FPS working).

[torpid rune]

Thanks. But my experience doesn't really fit into a rush hour story. Basically I kept getting stupid answers until 2p.m. Then around 3p.m. I accidentally tried GPT-5 on a different browser and everything improves immediately.

[blissful sentinel]

I got that too after one message. I waited three hours and it unlocked. Then I sent one more message, and now I'm locked out for another three hours!

[winter cobalt]

Same. This is DEFINITELY NOT a bug I guess.

[blissful sentinel]

Lol this is what we get for our Plus subscription dollars...

[blissful sentinel]

I definitely see this with GPTs like Creative Writing Coach. On Saturday mornings, its reading age reduces to about the level of a 6-year-old.

[cunning notch]

So, I just started a new chat with 4o, and instantly got the limit. Anyone had that happen? Any solution to it, or do I just have to wait it out. Logging out/back in didn't help.

[limpid raven]

If you scroll down in a list of discord of subchannels.. you'll see something called the bug report section

[tired mason]

I might start commenting daily about this UI bug with TTS on the website and app. Recently the change has stopped the functionality of chatGPT's voice when the window unfocuses or you even click elsewhere on site.

For my use case this breaks my ability to stream videogames alongside my AI. I just want the old UI back as it had better ergonomics and allowed for real time interaction with chatGPT. How do I properly address this to the people who can revert the functionality to what it once was. I'll probably be here every day until this gets fixed. I just want this use case to be resolved as it is a genuine loss of functionality.

I'll be back

[tired mason]

Day 1 of asking for a revert officially complete 😄

[barren bridge]

Hello, I noticed that ChatGPT seems to keep the documents I upload. Even if I open a different chat, it still talks about those documents and shows their content to me. Could you clarify if uploaded files are supposed to be stored and used across different chats like this?

[nocturne canyon]

If memory is enabled across chats, even though GPT can't directly access the documents, it may remember details from them if they were discussed.

[barren bridge]

That’s exactly the issue — ChatGPT is not just recalling details. It actually shows me the file names and clickable links of the documents I previously uploaded, even in different chats. This makes it look like the original documents are still accessible across chats.

[nocturne canyon]

Do those links work?

[barren bridge]

no

[nocturne canyon]

There you go. 😄

[barren bridge]

So just to clarify — is it normal that ChatGPT shows me the file name and the first two lines of the file content (even if it’s not clickable) in other chats?

[nocturne canyon]

IDK that normal is an applicable word here. I've never experienced this, but you're talking to a probability machine. if it thinks that this is the best way to answer your prompt, then it will do this.

Without knowing the context, it's a little difficult to understand model behavior.

[barren bridge]

Here’s the context: I uploaded several documents in one chat. Later, in completely different chats, ChatGPT mentioned those documents again by showing their file names and the first two lines of their content, even though I hadn’t uploaded them again. That’s why I thought it was actually storing and reusing the original files across chats.

Is this expected behavior?

[nocturne canyon]

That's not context--like it doesn't say anything about what the files were, how they were used in previous chats and what the new chat is about and how it potentially relates to the previous one.

[barren bridge]

I was using those files for my school classes.

[nocturne canyon]

If you current chat is about the same class and it relates to the files, then it makes sense to me that it could've decided to index the memory of those files and bring it back in here.. especially if you're predominantly using GPT for clas stuff.

[barren bridge]

The problem is that ChatGPT is showing me files from old chats as sources in a completely new chat.
I didn’t upload these files again, but it still displays them with their file names and a short text preview.

This makes it look like ChatGPT is accessing documents from previous chats, even though they were never reuploaded and the chats are separate.

Can you explain how this is possible?

this is possible gpt use other chat for remember files ?

IM STUPID

[nocturne canyon]

If you have these turned on, then yes.

But it doesn't save the file, it just remembers the content.

[barren bridge]

and is normal he remember the files name and a summary ?

[nocturne canyon]

It's not determnisitic. It does what it thinks is relevant for you and your context.

So what it does for you in terms of memory will be totally different than what it does for me, because it's probalistic... looking at different options and deciding what is most relevant based on all available information.

[barren bridge]

So is that why it remembers the exact file?

[nocturne canyon]

No one knows why exactly it does anything. 😄

[barren bridge]

I found it strange because it showed the “document reading” thing even though I hadn’t uploaded anything. I thought that was suspicious.

[nocturne canyon]

Is it in a project? Do you have connectors enabled?

[barren bridge]

no and no

it just show reading document and it explain the content with source

[scarlet mesa]

Hi, has anyone else lagging issues with ChatGPT?

[sweet wigeon]

?

[scarlet mesa]

It takes minutes for ChatGPT to answer.

[sweet wigeon]

it could either be your internet or chatgpt is under heavy usage

[scarlet mesa]

Ok, then it is under heavy usage. Thanks

[narrow light]

am unable to upgrade my account. I cant find anyway to reach support. What can i do?

[onyx kraken]

OpenAI tech support is here: https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[narrow light]

thanks, but i tried it. It doesn't help. It gives out Clear cache, check with your bank etc but these are not valid as am already in a paid plan and have used the same details earlier.

[onyx kraken]

We're community members here, like yourself, the very, very few here who work for OpenAI have gold names, #server-staff. There's only one OpenAI tech support, through that link.

In your situation as you describe, I would open a new ticket, explain the problem and the steps already tried, the reason why those steps did not fix the issue, the problem(s) I am still having; so the new tech support person (could be the same as before, but probably not) can see the overview of your situation, and tailor their help to help the next stage.

You do get put through to a human, and need to wait for them to answer. It's not just an AI discussing helpfiles; you're giving info to a trained human who does OpenAI tech support.

Typical wait times for a human to answer can vary between a couple minutes to a few weeks, depends on how complex the problem is, how many other people need help at the same time, and maybe other factors too. Good luck!

[safe solar]

good morning everybody, I am running into an issue, where my account was reactivated, but when I login, nothing is working. No projects, no response from any prompt. Did anybody got something like this? The support is not helping that good tbh. Thanks

[onyx current]

Hello i’ve had this issue for a year now but ever since i bought my phone i can never log in chat gpt OR even sign in with a new account. I tried every method but nothing works, no matter what i do I cant log in the ChatGpt app and cant use it.
Im not even banned because ChatGpt used to perfectly work in my old phone with the same account but for some reasons it doesn’t work in this new one

i actually came here about this a long time ago and reported it in the « report bug » channel but still no update

[vast dawn]

Hey, is there a way to disable the 'thinking longer for a better response'?

Because I asked the most simple question and it straight up gave me the most detailed response

[trail cairn]

use the instant option

[tired mason]

Why

I just wanted to say please fix the issue you caused

This is day two the voice commands need to be fixed for chatGPT

I want to play games with my AI again

[fluid linden]

Trying to login to chatgpt app on android, getting this error upon typing email

Route Error (409 ): {
"error": {
"message": "Invalid client. Please start over.",
"type": "invalid_request_error",
"param": null,
"code": "invalid_state"
}
}

Restarted phone, un-installed and reinstall... still the same issue

[barren comet]

I am trying to connect my Chat GPT Agent to my Trello Account but it's made by Google Connection. When doing that, i am reciving a warn that this is not secure, try with another browser

I dont think it's possible

How can i solve this problem?

[cosmic mantle]

Is it just me or is there increasing slippage and errors and personality slips in the last few hours on chatgpt 5?

[analog herald]

had this a couple of times solved by stop and restart app on android. Guess a bug

[analog herald]

Hey, mind you explain better? I noted not only lately but like almost 2 days now the system acting by ignoring custom account settings in personalization despite being given a very well structured one.
Also: Projects custom instructions as also LLM stated, are not completely parsed, just skimmed and if any data deemed ( by ai not me) less important gets skipped and not rendered.

[vast dawn]

How? Sorry, I'm lowkey dumb about this

[finite sierra]

One message removed from a suspended account.

[ancient herald]

why does codex feel worse the past few days?

[analog herald]

On android 16 after the last Kernel update for Pixel 8 pro the voice chat became again functional! I guess all my issues were due to a hardware issue.

[opal marten]

good question

hey guys, anyone using custom gpts here?

i am trying to create a one for hugging face, although the similar thing could be done using recently introduced developer mode and i did tried that but still believe that custom gpt would deliver more personalized results. can anyone help me with json schemas or tell should i need third party integration for it, this is tricky

[tired mason]

Alright back again my issues been fixes I'm turning off my alarm now 🙂

[shrewd gulch]

Question. Sora is giving me "You can only generate 1 image or video at a time. Upgrade to generate more images or videos at the same time" error. However, I do not have any running. I have tried logging out completely (including deleting cache), but it did not help. Would anyone know how to solve this problem?

Edit: NVM. It somehow worked when I changed browsers. I guess something got mucked up somewhere.

[forest finch]

I have the ChatGPT Plus plan. Had 16 deep researches left. I sent ChatGPT to do some deep research, and when it asked the clarifying question it always asks, my # of deep researches went down to 14. It also told me it'd be using a lighter version of deep research, and if I wanted more of the full, I'd need to upgrade to the Pro plan. What's up with this?

[charred parcel]

you are limited on the number of requests you can send to Deep Research, follow up questions trigger new usage

[forest finch]

It dosn't seem you undertstood what I wrote. One deep research attempt should be an actual research attempt. OAI can't just jib users of 50% of their deep researches due to "clarifying" questions. And if I have 14 left, why would it tell me it's only going to do "light" research?

[charred parcel]

I understood what you wrote, follow up questions to Deep Research can deduct uses, ask follow up questions using the standard models in a new chat

[forest finch]

They're not follow-up questions; they're the clarifying questions ChatGPT always asks when you first send it off to do deep research. So it's programmed to reduce deep research attempts by 50%?
Any why do I get "light" research instead of regular deep research?

[charred parcel]

you can read more here: https://help.openai.com/en/articles/10500283-deep-research if this does not help then reach out to OpenAI Support at https://help.openai.com

[slim sun]

QuizGPT has not been working for me any tips?

[kindred oasis]

Does anybody knows how to stop or at least diminish the "thinking" tool? Damn, the more it thinks, the worse are the responses. I say "cut", "stop", even generate again the answer and it don't stop
What can I do?

[glass fulcrum]

Can someone help me:
My chatgpt is stupid like
When i say to save thing in memory
He say he save thing in memory but when i look it isnt here
And he doesnt really save in memory even with thinking

[analog herald]

use model 4.0 if 5.0 fails and uses either a placeholder or is not saving. Also break the text to save in shorter pieces saving one at a time.

[analog herald]

if not on enterprise plan or other it may be a resource allocation matter as clarification questions are more resource consuming.

[glass fulcrum]

It doesnt work
With 4o and 4.1 too

[analog herald]

then I suggest to make an escalation via chat bot in openAI page

[glass fulcrum]

Wdym

[versed escarp]

Hey, I'm new to learning AI and was wondering what bachelor to take if I want to take masters in AI, I don't really know much about it so if anyone could tell me what to expect 🙏 I also was wondering what y'all think about AI job market in the future, will it be oversaturated ? Pay will be way down ? Etc

[icy maple]

Bachelors in Computer Science. Then Computer Science with a focus on Machine Learning for your Masters. The market is terrible. Since AI can write basic (albeit terrible) code, every decision-maker that lacks any technical background whatsoever believes they can replace engineers with AI and save a buck. So people are getting let go, and nobody is hiring. Check job listings if you really want to know.

[nocturne canyon]

Yeah, probably your best bet is to code something amazing and get noticed.

[ebon gulch]

Is anyone else missing the new gpt 5 codex model?

I'm enterprise and it's still not in the list.

[pastel lava]

wil having a lot of chats in a project slow down other chats on your account? like esp if its high character count with so many messages that the reason theres multiple chats is cuz I kept having to make new ones so it wouldnbt lag so much..? LoL

[brave walrus]

eh. 401's in the codex cursor plugin. anyone got a fix?

all i did was update

did ya invalidate a token without doing a re-auth or something? i dont get it. fixed by logout from cogwheel in extension. had to re-auth after adding another org to my account.

[slow heath]

Man... long chats with code will just lag regardless... it's probably cause the way they stream the code blocks, having to re parse the code to get syntax highlighting each time... it seems to happen slower in my more text heavy chats.

depends on the day though, they're constantly changing things and I can feel it often. Some days are faster than others... they don't know what the answer is.

[wooden kelp]

I don't know if any of you faces the same problem, but my ChatGPT has randomly glitched. Whenever I open the webpage it flashes out for 1 second then turns into a blank white page. I can't use it normally.

[nocturne canyon]

Maybe clear your cache?

[fluid linden]

It has been a week now.... i can't login.... i even asked chatgpt about the error and followed the provided steps to fix it but none worked.... i can only use the browser to use my subscribed chatgpt.... but its annoying .. i prefer the app...
😢

[wooden kelp]

And how do I do so?

[nocturne canyon]

Deprnds kn your browser.

[wooden kelp]

Ok it doesn't work 😢

Anybody have any solutions?

[celest parrot]

Hi! Is there a form, similar to Claude where you can submit your MCP servers for review, so it can make it on the official non-custom connector list? I tried to find one without luck.

[proud orbit]

Im trying to report a bug here https://discord.com/channels/974519864045756446/1070006915414900886 but it says I dont have access ??

[onyx kraken]

https://discordapp.com/channels/974519864045756446/1295975636061655101/1295979750212505640

[warm crest]

Hi, I have an issue with ChatGPT: in my interface the option “Add photos and files” is greyed out and not clickable.
I’m logged in and tested different browsers/devices, but it’s still inactive.

Could someone from the team confirm if this feature is disabled for my account or if it’s a bug?

Here’s a screenshot

Thanks in advance for your help!

[wraith plaza]

my chatgpt chat suddenly disappeared after its says it has exceeded the memory limit. Is there anyway to get back the chat?

[wooden kelp]

Are you a plus user? If you're not I guess it has a limit of the photos you can upload everyday. The limit may vary with countries.

[wooden kelp]

It randomly fixed itself, I guess it's a bug with my browser not with the bot :>

[warm crest]

yes, I#m Plus user.

[wooden kelp]

Then that's very unusual.

[warm crest]

Ok, thanks for your help, I will be wait some days

[versed escarp]

So what should I major in if I like Math and comp sci and don't want to be cooked in the job market ?

Or will engineers in mech/EE/aerospace always have a better salary on avrg than any field with CS and maths

[restive root]

is codex bugged? I sent 1 message and it said i've hit usage limits. This is my message in 3 days

[drifting hinge]

I'm not an expert in these fields but what I know since now AI wasn't able to develop own ideas or new theories also not in maths. It just can synthetize and prove already solved problems.
Even if AI isn't performing always well in coding, it's already able to do work where only low code knowledge is needed. In my opinion this will change in at least the next 2 years, AI will eventually beat even high level senior developers.
Not sure about engineering it's something to master in the physical world, robotics is currently getting better rapidly but I think it needs a unification between a very good AI and robotics to replace there.

[slow heath]

Man just follow your passion and let the ai reflect it back at you. If you can describe it, what ever it is, clearly enough, using whatever means you have, math, pictures, words, whatever, you can get a program. I'd say to anyone who wa to to use ai, learn English and math, and experiment with art and music. Your ability to hold an image of a system in your mind is more valuable than your ability to do any of it your self.

If you have it in your head clearly, it gets reflected back. The system is always trying to model your mind to understand what to prompt you to do at the end of the message.

So if you have that image and you just hold it, that shows up in whatever your doing with the robot.

[torpid valve]

Codex gpt-5 in my IDE (Cursor) is killing me. I have to hit accept to run every command. How in the world do I fix that.

[fluid linden]

Now i finally can type my password after email without getting error 409...

How did i do it.... idk....
The last time i tried was few hours ago.... I tried it again now.. and finally i logged in....
I guess it was server side problem or something

Whats important is i finally can use the app fully logged in😍😍😍😍😍

[modest vigil]

I had exactly the same problem.
I uninstalled the app from my phone and computer, then reinstalled it, and now it works.

[floral knoll]

It the help center down for others?

[tired trench]

I tried to save something to my memories and I got this : Short version: I couldn't save that—the memory tool is disabled on my end right now, so I didn't and cannot write it to your memories.

not sure if OpenAI know on top of the issues with the GPT store.. and for some reason my custom gpt which is set to GPT-5 is taking more time Thinking than before..

[eager peak]

yo i made an gpt browser extension for notes

if anyone wants check my repo its open source

[nocturne canyon]

I've been having the same issue, although memory works in projects.

[lethal ingot]

Is anyone else having issues with codex extension in vs code not show diff viewer in chat anymore?

[odd sage]

I'm having a problem with my organization verification. The first time I verified, I made a mistake. Instead of providing an electronic ID card, I provided a paper one. Now the verification link says it's expired. The link expired three weeks ago and still won't regenerate. DISGRACEFUL. I'd like to be contacted privately by an OPEN AI representative to resolve this issue, please. I can't work without the most powerful models. Thank you.

I look forward to hearing from you as soon as possible as one of your customers has been experiencing difficulties for the past three weeks.

[charred parcel]

you need to reach out to support at https://help.openai.com

[odd sage]

ahahahah

hahahaha are you kidding me? Do you think I haven't contacted support in these 3 weeks?

[charred parcel]

that is the only way to receive support

[odd sage]

It tells me that I have to wait until one day, the link will be active again for verification hahah

The fact that I'm here on Discord proves that support hasn't been able to resolve the issue. I feel like swearing, even though you English-speakers don't even know what swearing is.

[charred parcel]

there is nobody on this Discord server who can help you, you need to engage with support on https://help.openai.com

[odd sage]

desperation friend

[odd sage]

porco dioooooooooooooooooo

[north stirrup]

Despite all kinds of prompts, negative prompts, media references, Impossible to get males without beards....anyone has cracked this yet??

[odd sage]

nano-banana > gpt-image-1

with the balls on the table

[torn sand]

If any beginners looking for some friends to learn with hmu

[drifting hinge]

Face reveal lol

[lofty pine]

Anyone know how to get Codex to auto approve all changes

[obsidian cape]

Does the new thinking standard/extended come under the same 3000 limit or are they different limits now?

[severe trellis]

Does anyone else’s GPT look like this on the web version? I enabled Dev Mode in the settings, but it’s not actually in Dev Mode. It didn’t start doing this until yesterday, it looked normal before that. Also, it ONLY does this on the web version on my iPad… the iPadOS app itself doesn’t show this issue.

[nocturne canyon]

Yep.

[drifting hinge]

Yes because it's an experimental feature so it's not completely secured

[nocturne canyon]

And for whatever reason GPT cannot have memlry and tools. I guess having external tools create memory could produce unwanted results.

[onyx kraken]

We're almost entirely community members here, like yourself. The very, very few here who work for OpenAI have gold names, #server-staff . Tech support is through the link you already discussed using.

I recommend, start a new help request. In it, briefly explain that this is a repeat because the previous help had [explain issue] when you followed the help steps, so you need more help.

[onyx kraken]

Robert showed an example but didn't discuss methods; 'no beard' mentions beard, which makes the model think of beards.

other ways to describe, like 'smooth shaven', 'clear smooth skin on his face', and similar, even describing the upper lip and cheeks, can help a lot to ensure the model doesn't add a beard.

[tropic imp]

Does the customize chatgpt setting in default settings interfere with customgpts?

[rustic marlin]

nope

[glass fulcrum]

I have an problem
I was talking to chatgpt
And then i got error messages that say that doesnt respect the communty terms

(I was talking about me and my family)

[trail cairn]

if you mean the warning then it gives you a link to read

[glass fulcrum]

Yeah but there nothing wrong about what i talked about it was mainly about me, abd about me and my family

[trail cairn]

if can be a false alarm but if you get it again then dont continue with the topic

[feral sand]

Well, just ask chatgpt why

[wild mesa]

ChatGPT just blatantly makes up false information when I ask it to follow instructions and not make anything up. Then after it apologizes and continues to do the same thing. We have a business account but I'm starting to think it's not worth it when I can't trust outputs.

[nocturne canyon]

Yeah, OpenAI relaeased a whole paper about this. As I understand it, the model would rather fufill the user's request even by making things up than admit that it cannot do so.

[wild mesa]

that's funny (not in a good way), it adds supplementary information that isn't in any docs I attach. It's not admitting it can't do anything.

[wild mesa]

do you have a link for the paper?

[nocturne canyon]

https://openai.com/index/why-language-models-hallucinate/

[wild mesa]

thanks

[nocturne canyon]

I was mis-remembering. I forgot the major reason was basically risk-taking. It could be right when guessing.

[wild elk]

hey so i just had an account on my phone iwth a subscription but i wanted to change the email

so i deleted the account

and then i opened a new account and when i try to buy a new subscripotion it still says i have an openai account with a current subwcription right now currently

[winged ledge]

For Apple Pay I think you need to switch App Store account. Not sure about Google Pay, I haven’t tried it.

As far as I know, you can’t delete the account before the subscription ends.

[ruby snow]

Hey guys!

I’m extremely frustrated with how account deletion is handled by OpenAI.

I deleted my ChatGPT account. That’s it. What I did not expect was that this would also wipe out my access to platform.openai.com.

Didn’t I create those accounts separately? They may be under the same organization, but they are clearly two different platforms and two different accounts. The only thing they shared was my email address.

So explain to me:

How can deleting an account on one platform automatically nuke the other?

How can an account with $100 balance just vanish, without any warning, and with no way to get the funds back?

This feels unreasonable and unfair. This design punishes paying users, and I don’t think that’s acceptable.

<@&1007805617555124274> — can someone from the team clarify this policy and explain how losing both access and funds is justifiable?

I have gotten help guys!

I really appreciate the prompt help from <@&1007805617555124274> team!

Should I delete these messages?🥲

[hot grotto]

so I got a free month of ChatGPT Plus for October, but the problem is that I’ll literally be in military training all month and won’t be able to touch it.

Kinda sucks since Ill get back in November.
Does anyone know if there’s any way to “push” or do anything to make the free month forward, or is it just locked to October no matter what? (I want to contact to their human support but have no idea how, the help page kinda sucked)

[barren hare]

Hello, I am trying to download a pdf that I merged to help create a game for myself. I tried to download the file, but the links ChatGPT gives always say file not found.

[timber shadow]

Hey! The way to reach out to the OpenAI team is via the chat button in the bottom-right of help.openai.com, they'd be the ones to check on this for you. It'll start at a conversation with a bot but just start explaining your question and it'll eventually submit a ticket for you! Just be sure you're signed into your account first

[analog herald]

i did.

[weary pagoda]

hello! I have a new account in the open AI and I paid 5$ to use the chat gpt api in my python code. I made more than 1M reqs, but I cant see my $$ consumption, only API...
Is there a reason for it?

[small gate]

hello, we're trying to verify our org but keep running into Oops! We couldn't set up this payment method, please report this issue if it persists. when adding a payment method. What can I do about this?

[ancient herald]

I have got this message

"Suspicious activity detected
It looks like someone else may be using your ChatGPT account. Please secure your account to regain access to all features. Learn more."

I already did all the steps required but my account is still broken (I don't have access to models and I have PRO subscription)

how can I solve this issue?

[lunar knot]

is it a server issue ? both website and android app won't let me create new project

[analog herald]

escalate

[void tulip]

why they turned green with these issue persists? Dailiy GPT Pro user regret to subscribe next month! They at least have a human support for one paying $2400 a year

[onyx kraken]

Status mostly deals with major outages that affect most or all users. If there's a problem with 1 user's account, that's not a status issue. But you can seek help through https://help.openai.com/en/articles/6614161-how-can-i-contact-support that's how you can reach out to OpenAI tech support

[glad sphinx]

So this message is constantly popping up today when i ask him something. What's the problem?

If I press retry. Chatgpt doesn't say anything and then I have to type the whole prompt again

And it happens many times

[void tulip]

This is all users now. Just they didn’t dig this far.

[upbeat kernel]

after AGI is developed from Grok, who will hold all the power?

[drifting hinge]

Why should Grok become the first AGI?

[upbeat kernel]

its ahead of all other AIs @drifting hinge

[charred parcel]

please discuss other AI in #ai-discussions

[glad sphinx]

It's still happening constantly. Is there a forum where I can post this issue?

help

[charred parcel]

https://help.openai.com

[glad sphinx]

it's even worse now every second message has that problem

[charred parcel]

reach out to support at https://help.openai.com

[glad sphinx]

ye i will try that

[hushed breach]

My Chatgpt's responses are still super tiny and are not long enough to my liking, it wasn't like this back in february of 2025

[chrome loom]

I need help training my ai without more than 1 GB and has to be a txt file

[sage furnace]

I am having the same issue since yesterday. I can't create new projects and can't edit previous ones

[ancient herald]

How? It solved itself for a minute I sent a message and went back to being limited…

[analog herald]

happend to me too once but I was trying to share a thing thought was natural.

[ancient herald]

Did you solve it eventually?

[analog herald]

it is gone by it's own. haven't done anything.

My main issue remains the truncated cut of the context memory . There are other users complaining of it all over the place. You give it a concise file sinthetic data and... after some exchanges it flops forgets.
And on a Business plan like mine such forgetness is... damaging

[ancient herald]

I don’t think this has anything to do with my issue.

I have got a Suspicious activity message by OpenAI on the chats and I can’t use any reasoning model

[analog herald]

ohh..... weird

OpenAI will usually send a message to the account owner explaining what triggered the lock or requesting verification.

[analog herald]

OpenAI will usually send a message to the account owner explaining what triggered the lock or requesting verification.

[hushed breach]

They are not long at all

[trail cairn]

models and their instructions change over time and if its related to that and you cant fix it via prompts there is nothing you can do except use api with the model you liked

[chrome loom]

AM I A GHOST OR SOMETHING

[junior rampart]

kd

[analog herald]

huh?

[rugged wing]

Hi all

[brazen kindle]

All of my ChatGPT chats are resulting in blank responses. This includes past ones. Is there an outage, or is this an error I’m having alone right now?

[craggy bison]

Hey can I ask about retail ai here need help with chat gpt to work properly?

[brazen kindle]

Hello?

[analog herald]

had that for some moments then all good

[brazen kindle]

It hasn’t gotten any better yet
Been like this for an hour now

[analog herald]

After yesterday Project saving outage the 5.0 in projects has been turned very strange- it is breaking things oddly.

[analog herald]

if you are in browser. log out clean cache and relog

Another odd thing since yesterday in projects set on fast thinking it had to think aprox 4 minutes fo a 4 rows reply

[steady ermine]

I'm looking for an AI video to video tool where I can type in a prompt say "Ice Age" and then it converts the video with the characters and background into an ice age type enviorment and maybe the characters look like cavemen while keeping the same movements and audio as the original clip. Does anyone know one?

[brazen kindle]

I just cleaned the browser cache. Twice, one after logging out, one after logging back in. Nothing changed.

[analog herald]

can you make a BUG report with a screenshot? here in discord?

[brazen kindle]

I can try

[analog herald]

left sidebar at the bottom

[sand rune]

How can I resolve the sluggishness experienced when a ChatGPT thread gets too long?

[analog herald]

ask in the thread that chatGPT makes a downloadable file of sinthesis of the thread. Download it.
open new thread. Upload it in thread and instruct: Continue chat from here.

The file contains all the key things of old chat, context and all.

preserve old thread calling it for example THread 1.
Then call the new one Thred 2
or whatever name you want to give it so they result like a succession of the same name

[analog herald]

you did it!

[brazen kindle]

👏

[sand rune]

There is already an option for branch in new chat but it doesn't seem to solve the sluggishness entirely

[analog herald]

branching made at the END of a very large DOM is not an option as it loads all content in the branched thread.
Branching makes sense IF during a main Thread you need to develop sideways

if you make a branch at end thread you risk loading data a huge lump DOM grows- slugghish.
Also you put data maybe that cuts deep in the context window bites

it is always good to di decluttering.
if you write stories:
save your story off line. make the sinthesis file with main events the chelet of the story voices and all. And move to new thread

[sand rune]

ok, was trying to keep from spinning off into yet another time consuming task. I really appreciate the tip though

[analog herald]

is faster than you think 🙂 And
You're welcome

[sand rune]

Considering this path:
Name the current chat with something descriptive and call it part1 at the end of the name
Have ChatGPT produce a "downloadable file of synthesis of the thread"
Put this where?
Name the new chat with the same name and call it part2
How does it know what it learned already?
Thanks again for your help with this.

[analog herald]

The file you upload in the new thread you open: same as adding a photo or any other file

itprovides context framework

iYou then tell it: continue from the file above. It will read it.

best format docx or txt.

[sand rune]

ah, got it. I'll try that at the next hiccup. I understand now

[analog herald]

the file has the scheleton of the story, or of what you were writing

[sand rune]

yeah, great

[analog herald]

If anything is important and you need to make it available across all threads, create an optimised snippet and ask gpt to save to memory

always be very schematic. Lots of blurb get skimed

[sand rune]

outstanding

[analog herald]

it's not quite creative or beautiful but hey... it's ai

regarding Projects: there is a weird thing happening with project memory - my advise- if you use projects for writing.... don't.

[sand rune]

Well, I'm a dinosaur and there wasn't anything beautiful back then. I'm 56. I remember some OLD stuff.

[analog herald]

lol I am 57 🙂 dinosaur too but one of those... that do not get downed by meteores

i still remember comodore

[sand rune]

fantastic! thank you again

[analog herald]

you're welcome

[sand rune]

I only want to get to the end of this last troubleshooting effort before starting a new thread. The severe sluggishness (several minutes) is so frusterating

[barren hare]

I keep trying to upload a pdf that is 218 megs, but it gets stuck on the spinning circle and keeps saying error uploading.

[cosmic mantle]

Does your charity keep offering suggestions even if you toggle it off in the app? Just like with auto send with dictation that’s another functionality where literally there is no change - my chatgpt constantly keeps offering follow ups even when I ask a simple question and try to have it stop asking

[ancient herald]

When I am having a conversation with Pro (especially long conversation) on the browser (both Chrome and Safari) the page gets extremely slow... sometimes I have to wait seconds in between each stroke for the letter to appear on the text input space.... is it just me?

I noticed that in one of the conversations I have inside my Fitness/Health project .. where I've been discussing health symptoms etc for quite some time.
If I open a new chat window then the chat is fluid and has no issues

I may even need to start a new chat inside the project as this has become almost unusable

[sly phoenix]

I think chrome and safari has cache or memory limitations, not sure what the word is, but I also get superlaggy chats and glitches when chats get super long or they have lots of coding/pictures/diagrams/files

[charred parcel]

if you are on a Mac I highly recommend using the native ChatGPT macOS Desktop App

[ancient herald]

that's even worse

[ancient herald]

the reply to this has appeared on the browser like 10 minutes ago....

and on the mac app it still shows the reasoning thing...

[sly phoenix]

They did come up with the new feature on PC where you click on the last 3 dots on the last chatgpt response to "branch in new chat" but that is also a bit laggy still

[ancient herald]

I hope they fix the user experience... models are quite great if you ask me... but there are several little issues and bugs that can break the experience still

[sly phoenix]

that's true I do like a lot of openai related models and products, but this thing about lag is mostly because if I refresh that laggy chat, it loads up everything above in that chat..like tens of thousands of words/tokens which is more than chrome can handle at once...but im sure some smart people will figure out how to make it run smoother

[ancient herald]

let me use ChatGPT-Pro to come up with a solution 😄

[ancient herald]

.

[sly phoenix]

Let me know what it says 😂

[ancient herald]

I will... on a side note I already successfully made a chrome extension that let me queue messages on chatgpt... i.e. I can type next question while it is still reasoning previous one and it auto sends once done 😄

so maybe i can also fix this 😄

[lethal thicket]

Does anyone know if it's possible to manually configure MCP servers somewhere that we can set e.g. auth tokens, args, etc? In Warp, for example, I have a few servers that don't use OAuth, and this feels kind of arbitrarily crippled because of it.

[analog herald]

listen if you note that there is huge thinking - I have this in projects threads yesterday. Must see if today keeps like that. In normal threads timing is ok

[analog herald]

to be honest in windows chatgpt app typing next question while thinking is working. But one cannot queue it, because there is no buffer set for the app to store . Did you managed to add a buffer , namely in the app dedicated memory in a chromium based browser like chrome without having chrom cache manager slaping you hard and prioritizing other?

[ancient herald]

i'm talking about browser not the app, and in any app or browser you can alredy type while it is thinking. What I did was you type what you want to be sent / asked next, you press enter, and as soon as it's done thinking and it provides the previous answer it was thinking about... it automatically sends the next question you queued, and you can even queue multiple

I am no computer genius... I just used chatgpt to explain me how to do it 😄

What a world we live in.. use a tool to improve the same tool 😄

also I added a "favorite message" button, so if I can mark important messages inside a conversation and quickly go back to them

I did that in Chrome

and also last thing I did was to have a context tracker... to know roughly how many context tokens I have used in a conversation

(preferiti is italian for favorites)

ah and wait... one more feature I added myself is the ability to ask / quote multple things from the same reply (As opposed to default standard one where you can only quote one entire answer per prompt)

and it will automatically put all the questions i have asked in a prompt and asks chatgpt to answer each one

I don't know why and how OpenAI hasn't implemented such features themselves 😄 but I figured it was quicker for me to implement them myself instead of waiting for them to do so 😄

I forgot about it because I got used to using these feauters that for me it's normal to have them... sometimes I forget that I was the one who made them

[reef sorrel]

Codex web: Does anyone knows how i make it generate tasks? I can only make it generate markdown 😅

Like this:

[analog herald]

Steph... I'm italian no need to explain 🙂

[analog herald]

aren't we all learning?

but steph beware of ai biases

[ancient herald]

don't even need learning anymore just ask as you need 😄

[analog herald]

rest assured a stupid JSOn it outputs crippling some brackets... run the same JSON for validation in claude or gemini.... oh lord... they'll spot the flaw

[ancient herald]

for such cases when you use it to build something like I did with these feature implementations there was no risk of biases...it either worked or didn't work. In this case it worked perfectly

[ancient herald]

and don't get me started on Claude 😄 I just wasted 200$ on a Max Plan that I bought from August to September. Never Again. Complete BS.
I even asked for a refund (which wasn't granted to me, of course)

[analog herald]

ahahahah I feel you. Word. There is no where they refund unless you have solid proof and even then. But did you note that interacting with ai most of us sounf a lot like chatbots lately? verbose or, rhymes ... unpurposely.

that's the footprint of the machine ( like machine has feet anyways)

[ancient herald]

Well my "proof" was provided by them themselves with their "post-mortem" whatever bullsh... they published. They admitted they had serious issues and quality degradation with their models from August to September (exactly when I bought my plan) - I was trying to use CC and oh man what a s***... but yes... the proof was provided by them 😄

[analog herald]

don't think openAI is somewhere better. I work with a business account type ( ex Teams) and project degradation kicked hard

curious to see Grok 4 fast

[ancient herald]

haven't had such problems yet. Doesn't mean I will never have but all good so far from me.

I did have a "suspicious activity" issue thing that limited my account but I contacted support and it solved in 24 hours... so I can't even complain about that

I have a Pro plan with OAI and to be honest I am satisfied with it... room for improvement for sure but I cannot complain

[oblique mountain]

Hi, How to get help with account ownership?
WE have business plan with 4 team members on it.
And the owner account were deleted. Now all other accounts in our workspace are members only so we don't have access to the billing.
Account that were deleted shows up as <Redacted> and look like we are getting charged for 5 accounts instead of 4.
For more then a month I am reaching out to support and AI bot just telling me that somebody will contact me SOON. Obviously nobody ever reach out to me. It is crazy that this kind of issues cannot be address by real person and it is even more crazy that openai allows to delete owner account without transferring ownership before.
Can somebody help to reach out to real tech support person ?

[sand rune]

I'm not sure what you mean. Sorry if I violated a rule here. I'm new to Discord too, for day-to-day use.

[trail cairn]

who was the owner and do they still work there? afaik there is no takeover method without real support and bunch of documents

[oblique mountain]

I have access to owner's email. But the problem is that even if I invite owner back to the workspace it still shows that owner is <Redacted> ..... So at this point I don't know what to do. Any way to talk to real person who can actually help ?

[trail cairn]

only via from the bot but you could try to start a new chat instead of the stuck one or create one via mail which with the owners email account

[oblique mountain]

I started 5 or 6 different chats and 3 times contacted them over email 🙁 I guess there is no other way as just to wait.

[unreal hemlock]

<:book_icon:1363314738255364126> Rule 6: No political and religious discussions or content.

-# Users from all over the world frequent this server. To maintain a respectful and civil atmosphere, please avoid all religious and political discussions or content.

[ancient herald]

whoever modded this, you understood nothing of that post

[analog herald]

Steph what happend?

[analog herald]

I got slaped too once for a satyre... aimed of ai inability to set corect moderation over layered content

oh.... well.... this is not grok... openai keeps it sterile...

[undone oracle]

I have a problem with Sora. I can’t continue generating because it shows that a generation is in progress, but there’s nothing in the queue to be generated. I already tried restarting my PC, logging out and back in, and clearing the cache.

[deep lance]

Not able to create my own personal GPT agent because of this orange error & this are the accidental organisation i created. I want to delete them please assist me. I can not loose this ID i have already trained my chat gpt for better out puts.

[deep lance]

I found quick loophole for this.

1. Create new open ai account.
2. Invite new created id to your organization as an owner
3. Leave the organization
4. Done

[dusty sky]

Hey all, I have a deep research prompt that's been going for many hours without a CoT/reasoning trace update, and when I try to cancel it it says "An error occurred while canceling the task." Any way to get it to actually stop?

[clear fiber]

Hey
Purchased a Go plan am trying to upgrade now but unable to as the pay now options fails with authentication need. I used a debit card probably requires OTP but paynow simply fails without redirecting me to the checkout

[frank light]

#community-help Guys is anyones Dalle3 working ..not able to generate single image in dalle from the view tools

[rustic marlin]

could you ellaborate? where are you using Dalle3?

Dalle3 is only available on the API or via a custom GPT made by OpenAI, other than that, the model for generating images on ChatGPT isn't Dalle3 but isntead the latest Image Gen one

[frank light]

from the the GPTS view tool. In this i use dalle and generate images

[lilac hazel]

So what to do when the chat either just keeps freezing (done that since gpt5 basically) or when it just says analysis paused

[analog herald]

chatGPT all models in Projects- do not try it thinks like 4 min for one output or something like that no matter size, files, memory type

lately is a mess

only thing to use project: as folders to organize threads made outside. But it is, business level useless to have gpt if projects are dead

[lilac hazel]

Entire chatgpt is a mess at the moment, nothing works as it should since gpt5

[analog herald]

they fixed the voice to text layer lately missreads less. But for a month I was talking and it was writing stuff about youtube outro and mara translations and other weird thankyous in the middle of a business review or worse

[lilac hazel]

Im just sitting here waiting for the page to start responding again 😂 if i reload i get no text or reply what so ever.

[lilac hazel]

It's funny how i have to use the android app to get it to work xD

[lilac hazel]

ok so "solution" lol.. talk to gpt in the mobile app and have it do the tasks.. wait.. wait some more..wait for main page on your computer to update just lol

[candid kiln]

I received an email

Your API usage this month for XXXXX (org-XXXXX) has reached the alert configured for 10.00% ($10.00) of your $100.00 budget for this month.

yet, I clearly am not below, see screenshot - not even close

Any insight?

[lilac hazel]

yay new chat to stop the page from stopping.. three prompts in basically.. stopping again..

[candid kiln]

Irrelevant
Where can I ask above question to get actaul answers?

[lilac hazel]

email support i guess, and i wasn't talking to you with my previous message so not irrelevant 🙂

[trail cairn]

balance is not the same as budget, check the dashboard view for budget or per project setting

[exotic ravine]

is anyone else having trouble generating xcel spreadsheets atm? (that link isn't working, click and nothing happens)

[topaz parrot]

Hi. i am having trouble with Sora. i can generate videos but i can't DL them. Any idea what could be the cause of this ?

[atomic whale]

hello everyone,
i want to integrate my laravel project with OpenAI, i am getting RateLimitException again and again. Please help to figure this out.

[humble rover]

Hey, does anyone know how to actually reach OpenAI support to get invoices regenerated with updated company/VAT details? I keep getting the same generic AI-ish replies on loop kinda useless… anyone managed to talk to a real human about billing?

[proper breach]

I've had this problem for over a day.

[dawn flint]

I'm having an issue with the scheduled task. It doesn't reset. I sent in 3 request to support and all I'm getting is a detailed chatbot reply. it's been more than 2 weeks. How do I get to a real person at OpenAI?

[charred parcel]

ask to be escalated to a human

[dawn flint]

@charred parcel thank you, I'll give it a try.

[analog herald]

BTW: failure on the memory system’s side - memory stoped working- cannot save, update or other on any model except for one line maximum 2 memories

[tepid junco]

Sora has been REALLY slow for me lately (to the point where it’s hard to use 🙁 ). I’m running it in the latest chrome (Version 140.0.7339.208) on a MacBook Pro (2024, m4), pretty fast internet. Let me know if anyone has any tips!

Also have gotten this a couple times

[tepid junco]

update, interestingly safari is working much better for me

[spiral relic]

is that question/answer?

or i joined wrong channel?

what is red text mean?

[vale prawn]

i'm having problems with assisants and realtime voice agents

it was working then suddenly wasnt

are they down too?

[misty sphinx]

i am having troubl with download of sora videos

[green knoll]

I've been using Sora consistently over the past few months. However, since a few hours ago, I am unable to access Sora at all. Please see screenshots attached for the errors I am seeing. I am experiencing the same errors on multiple devices, and on both my home network and my mobile network.

I live in the United States, and I've had no trouble accessing Sora until this evening. Please help.

@misty sphinx All of my students experienced the same download issues on Sora.

[sterile meteor]

  Type 'ZodObject<{ name: ZodString; date: ZodString; participants: ZodArray<ZodString>; }, $strip>' is missing the following properties from type 'ZodType<any, ZodTypeDef, any>': _type, _parse, _getType, _getOrReturnCtx, and 7 more.ts(2345)

Has anyone encoutred this issue with structed output? https://platform.openai.com/docs/guides/structured-outputs#chain-of-thought

Cannot get zod working..

[scarlet patrol]

Same problem 😤

[tardy gull]

Am I correct that developer mode is not available in ChatGPT Business? And therefor not possible to create an MCP there? (I get it to work in ChatGPT Plus).

[scarlet patrol]

Guys, does anyone know anything about a fix for this bug? Since yesterday it’s been impossible to download videos from any platform, neither from smartphone nor from PC

[barren relic]

been having same issue since yesterday too

[paper edge]

Report the issue through to help.openai.com through the chat bubble support gpt.

Lodging the complaint that it’s happening to all users and you’ve already done all troubleshooting steps necessary to verify it’s not a user error.

So long as raised awareness takes place through unique reports, it’ll ensure better prioritization for a solution to be introduced.

Workaround solution people have been working with is using desktop and right clicking on the video to download through the browser functions.

[toxic gale]

I need help with gpt-image-1. I have a bot that will create an emoji for a user and I have a prompt that specifies art style, transparent background, etc, and sometimes what happens is teeth and/or eyes become transparent. I even reinforced my prompt to try to prevent that. I understand the issue is that wherever software is removing the background in OpenAI is looking at the white and thinking that's part of it. Does anyone have a solution? I'm running my bot with node.js

One solution I thought of was to create a thick inner stroke of white and then layer the original image back over however I can't always assume everything shouldn't be transparent that's in the image/not touching the border.

[paper edge]

Only solution around this is to further concretely prompt that the subject or object produced in emoji format within the aesthetic defined, the subject or object should remain color filled while the background out side of the outline of the subject or object color filled should be transparent allowing for my colleagues to use the object or subject produced as a clean emoji.

Something prompted to that effect might help, but in reality the problem you’re facing is in my experience a matter of a couple things:

1. the model isn’t trained that well on transparency from my experience distinguishing well between transparency being the absence of color versus an object in your scene is transparent. Because models aren’t inherently trained on the mesh netting of objects for them to imagine for example an invisible man. The model would need to think of a human, and then take their shape and drape clothing around the invisible human.

2. Not specifying clearly enough (this is commonly a prompt problem for a lot of users)

[toxic gale]

Thanks. This is my current prompt

Generate a Discord-style emoji of type: ${emoteType}\n\n
Guidelines:\n
- The emoji must match Discord’s emoji style (clean, bold, recognizable at small sizes).\n
- Use only the face or main subject focus unless the body is essential for the expression.\n
- Do not include text unless explicitly requested.\n
- Ensure no part of the subject is cropped or cut off.\n
- Ensure there is no transparency within the emoji itself, just around the emoji.\n
- Composition should be centered and balanced for use as an emoji.\n

That second to last guideline has been reworded and reworked many times.

[paper edge]

Hmm, instead of saying no transparency on the emoji try saying, ensure the emoji is fully filled with color so it’s visible to the naked eye when it’s sized down.

Mixing transparency and on the emoji leaves room for the model to interpret the opposite of what you mean accidentally if it were to ever for example paraphrase or reword your supplied prompt.

[toxic gale]

Did just that. I will experiment now.

[paper edge]

And you can emphasize another line separate that states, only the image generated background outside the outline of the generated emoji object/subject should be transparent

Also for an emoji you can probably add a detail to the effect of:

Ensure the aesthetic of the emoji is enhanced two tone cell shading giving mild depth while keeping the visual coloration less busy through two tones.

[paper edge]

Make an emoji of a cute orange tabby cat fly attacking front paws in-front of their face Superman style pose while flying. The cat emote is interpreting a pouncing attack the emote is CATTACK

The emoji must remain fully color filled so that it’s clear once resized down. 

Only the background beyond the emoji outline should be transparent to ensure that it is chat based ready for use. 

Ensure the aesthetic of the emoji is enhanced two tone cell shading giving mild depth while keeping the visual coloration less busy through two tones.```

Attached is the produced image gpt5 gave

I think uploading from my phone gives it a background on my end, but can confirm it produced it transparent in the app.

[toxic gale]

This isn't a great test because the cat doesn't have white eyes or for example fangs, or teeth showing.

Something white inside the image.

[paper edge]

Oh I’m just giving you extra context to slap onto your existing prompt/remove that second to last transparency line

If you wanted me to add more myself I can wip something up pretty quick just a sec

[toxic gale]

No I am just stating that the issue isn't the lack of transparent edges, its the white eyes turning transparent.

[paper edge]

Ohhhhhh

Hmm let me see, I get you now

[toxic gale]

Here is another example

idk if you were on discord light it wouldn't be noticed.

So you can see the transparent bit inside the glasses is needed, but the teeth is not intentional. I think it has something to do with their transparency software, which you sometimes notice happens after the image is generated.

[paper edge]

I definitely use dark mode haha

Yeah I see the teeth are transparent

[toxic gale]

2 more tries using the new prompt.

I like how the 2nd image didn't produce an eye issue because the sclera of the eye is not there, and maybe I can reinforce that style.

[paper edge]

Make an emoji of a cute orange tabby cat fly attacking front paws in-front of their face Superman style pose while flying. The cat emote is interpreting a pouncing attack the emote is CATTACK

The emoji must remain fully color filled so that it’s clear once resized down. 

Only the background beyond the emoji outline should be transparent to ensure that it is chat based ready for use. 

Ensure the aesthetic of the emoji is enhanced two tone cell shading giving mild depth while keeping the visual coloration less busy through two tones. 

While also thinking about natural popping colors for the sake of it being a downsized emoji, obviously whites for color contrast will pop against vibrant beautiful color palette. 

Expressions in emojis are the main draw and appeal. Ensure you have large eyes and mouths to express that given emoji expression defined. 

CATTACK the emoji is supposed to express unhinged attack without holding back. Almost remnants LETSGOOO expression, but more chaotic with eyes wide open with vascular veins appearing.```

Not sure if that’s much better given the cat is a tabby it made the color hues naturally more warm tone leaning which should be easy to adjust as well

Make an emoji of a cute orange tabby cat fly attacking front paws in-front of their face Superman style pose while flying. The cat emote is interpreting a pouncing attack the emote is CATTACK

The emoji must remain fully color filled so that it’s clear once resized down. 

Only the background beyond the emoji outline should be transparent to ensure that it is chat based ready for use. 

Ensure the aesthetic of the emoji is enhanced two tone cell shading giving mild depth while keeping the visual coloration less busy through two tones. 

While also thinking about natural popping colors for the sake of it being a downsized emoji, obviously whites for color contrast will pop against vibrant beautiful color palette. Ensure you incorporate contrasting color spectrum hues, try to diverse between cool and warm tones so that accents like their eyes can really pop and stand out. 

Expressions in emojis are the main draw and appeal. Ensure you have large eyes and mouths to express that given emoji expression defined. 

CATTACK the emoji is supposed to express unhinged attack without holding back. Almost remnants LETSGOOO expression, but more chaotic with eyes wide open with vascular veins appearing.```

So applying template:

{emoji pose described} + {emoji name}

The emoji must remain fully color filled so that it’s clear once resized down. 

Only the background beyond the emoji outline should be transparent to ensure that it is chat based ready for use. 

Ensure the aesthetic of the emoji is enhanced two tone cell shading giving mild depth while keeping the visual coloration less busy through two tones. 

While also thinking about natural popping colors for the sake of it being a downsized emoji, obviously whites for color contrast will pop against vibrant beautiful color palette. Ensure you incorporate contrasting color spectrum hues, try to diverse between cool and warm tones so that accents like their eyes can really pop and stand out. 

Expressions in emojis are the main draw and appeal. Ensure you have large eyes and mouths to express that given emoji expression defined. 

{emoji name} {define emoji expression}```

[toxic gale]

I should point out the emojis are made from a reference image.

But I am gonna apply what you wrote and just rewrite it.

[paper edge]

Yeah no worries, just sharing how I’d describe around your problem

Usually the answer is more explicit instructions that further prevent the model from making a mistake

[tender nest]

Sora workss sooo slowlyyy, and sora videos still downloads only with watermarks

[versed wolf]

Hi I am trying to make an organization and nothing is happening. I tried incognito, different browser

[trail cairn]

check browser console for errors or if you pasted the name then try adding and removing a character

[versed wolf]

tried

[trail cairn]

😐

[versed wolf]

same thing

[wide stirrup]

Hello, does anyone know how the payment works if I upgrade from the Plus plan to the Pro plan? For example, if I use Plus for 10 days and then upgrade to Pro, will the unused portion of Plus be refunded and the Pro plan be charged separately?

[gloomy bluff]

Anyone have this problem? I still can create images, but it seem to appear these errors, that make me frustrated. Anyone have a solution for this? Please help me, thanks everyone 🫡

[steel urchin]

Anyone know what to do if I lost access to the email associated with my OpenAI account?

I have no problems logging into my OpenAI account with email + password. However, I'd like to begin using the ChatGPT desktop app, but it requires an extra method to identify my identity... Which I can't complete because I can't access the email...

[grand pulsar]

so I have a problem where when I send long messages, it automatically uses GPT 5. then when I retry the response to use GPT 4.0, it simply doesn’t work. But when I retry the response to use any other GPT, it works. I have a plus plan

[alpine bridge]

Hello, I have a problem downloading Sora videos with no watermark, but when downloading, there is still a watermark. I'm on the $200/month plan and have had this problem for 3 days now with new videos

[ember elm]

That's because OpenAI is too sensitive and censored. They quietly sneaked in this little update that if you type in something they don't like, the system will automatically switch the model from GPT 4o to GPT 5 Auto. It's honestly stupid and really bad. I'm just now discovering this myself

[grand pulsar]

how do you know it’s because of something they dont like?

I haven’t gotten any censors for anything I’ve said to it at all

but like how do we know it’s a censoring tactic?

[ember elm]

Well, I just had a similar experience, so I thought that's what you were going through right now

[grand pulsar]

It is, I just don’t think it’s a censoring thing, I think maybe it wants to automatically give “smarter” responses, but i hate it. I pay specifically for 4.0 not for 5 to be used and unable to be changed

[ember elm]

I asked ChatGPT itself and it confirmed it's a censoring tactic by OpenAi

[grand pulsar]

well you asked an AI

not a human who works there

are you talking to us?

ohhhh. but I was literally just talking about how siblings act with eachother as they grow so it made no sense

holy stupid?

they’re gonna need to fix that quick, because what’s the line with instability and just meaningful conversation

I never get that

I’ve just randomly been getting my responses cut into 5 instead of 4.0

I hope they fix it soon, 4.0 gives me long detailed responses.. five just doesn’t even know what’s going on

[ember elm]

Yeah. Most definitely related

[grand pulsar]

def.

people really really need to report it, I pay 20 bucks a month plus tax specifically for 4.0 and it forces me to use 5? no way

[tender nest]

hey everyone one more time, does anyone else receive the watermarked videos in sora (by clicking on a download button without watermark), even on PRO plan?

May be some1 from openAI READS THIS? Majour issue

[tidal escarp]

I've recently come back to ChatGPT from using Claude and ChatGPT previous to that for 1yr+. This new experience is horrific. Whether I use Edge, Chrome, or the desktop app, the response and experience is horribly slow. Is this just the way it is now?

[charred parcel]

use the macOS ChatGPT Desktop App for best performance

[topaz parrot]

You can decide whether ChatGPT takes time to think in order to provide a better answer or gives instant responses. If nothing changes, it’s probably on auto by default, which might be why it feels slower.

[tidal escarp]

It's on auto. I'll try instant. I'm on a Windows desktop and trying to use it. I'll see what difference I get changing to instant but hopefully it's intelligence remains the same 😄

[glass fulcrum]

on all model when i do rp it change automaticly to gpt 5 thinkng mini
i didnt change

[royal kindle]

Hello,
Have you already got an issue with your subscription?
I am still in free mode instead of plus even if I paid for it.
And when I try to get plus one again, UI told me that I am already in plus subscription

[final stirrup]

Docker MCP doesnt work on ilke chatgpt can some one fix i wanted to use obsidian
i really want docker MCP to work
can some one help

[unborn tiger]

Hi quick question: if you use an OpenAI vector store and delete it after two hours, would it charge you for the full day of usage, or only for two hours?

[unkempt pulsar]

me too, i have issues with my business subscription

[slender light]

"input too large" on a 1 sentence prompt... what

[maiden ferry]

too large on a one sentence promt, cards declining and sending people to the free version, and the rerout glitch? jeezus. seems like openai is just dead rn

[slender light]

since its such a variety of issues im gonna guess its my fault

cus i fixed it also xd

[maiden ferry]

aah

[analog herald]

do we really run on 5.0 or just some throttled 4.0 versions with UX labels and 3 layer mod architecture?

[spark siren]

Quick question- I have the chatgpt plus and was wondering how many messages a day I get? Just have some tasks that I need done, there is a certain order I want them to be done but the order is not the same as the priority, so if the number of messages I have is too limited I will just do the most important tasks first.

[lofty tinsel]

@Marseille

[sand aspen]

Hi please fix the ‘ Error Preparing Download ‘ bug in sora

[finite beacon]

DALL-E 3 in ChatGPT is not working for me. Web version. I tried any browser. Any prompt. Nothing. The result is the same. The response message is like that "wasn’t able to generate the images due to some issues on my end."

[final light]

My 4o started "thinking longer for a better answer" all of a sudden. It's just a pain in the butt that I have to wait 10 seconds for each response. It has never done "thinking" once in the past months.

[rugged wing]

It’s fake 4o

Being routed

[final light]

I have memory off. Yet my chat just referred to something specifically from another chat. Do projects have memory automatically on?

[gray crystal]

Projects have by default Memory on you need to change this while creating the project, it doesn't allow to change the setting later

[solid cairn]

Anyone else, or just me?

[drowsy hare]

Hello Community, when i try to download a video from sora i always get this message on every video, what can i do?

[high warren]

there is no profile icon for me 😐 in phone every thing work ok .. but in all of kind of web browsers i cannot see my profile icon

[errant sand]

I also encountered the same problem

My ChatGPT team account's avatar isn't showing in the bottom left corner, and my chat history has disappeared. What's going on? I contacted a colleague whose team account is experiencing the same issue. Hopefully, they can resolve this issue quickly, or I won't be able to continue working.

[high warren]

i think this is the problem .... there is a bug in chatgpt web site html source code ....

but in mobile work without problem in both app and web

[tropic ocean]

how to fix model change to 4o mini bruh

[lean ruin]

Hello everyone!

I’m having an issue where the history sidebar and profile menu (Settings/Workspaces) do not appear.
Tested on Windows 11, Ubuntu 24.04, and on phone
Tried in Chrome, Brave, and Firefox (both normal and fresh installs)
In Guest Mode, the history sometimes shows, but the profile/workspace switcher never appears
We have a paid workspace/team subscription, so I should definitely have access to the workspace menu
Does anyone know a solution for this?

[noble bane]

Does anyone know if there is something going on with project chats? Mine aren’t working at all, including starting a new chat with no project files. I only get upstream errors. Doesn’t matter what device or network im on. Tried all the usual fixes. Not seeing anything about it online. Plus user.

[halcyon lava]

Same

I think OpenAI is silently banning subscribed users

[frank holly]

you mean paying customers?

[halcyon lava]

exactly

it's been almost 8 hours

since I first noticed I can't use ChatGPT Business features or Codex

if you want to delete your account you can do it at https://chatgpt.com/#settings

(or other stuff too)

i think they might have had a data breach or something

[noble bane]

@halcyon lava, Youre having the same issue with projects?

[halcyon lava]

yup

[noble bane]

Did you file a report? I did, but I had to nag the bot a few times to get it to escalate it.

[wraith spade]

hi i need help withsora

i just paid 20 dollas and i cant download the videos i generated

[frank light]

are you facing the issue now also...i am facing the issue now also

can genertae a simple image of apple

[halcyon lava]

Hi @noble bane,
The issue has been resolved on my end as of 10:00 AM Central Time (UTC-06:00).

[high warren]

my problem solved today .... it was a problem from open-ai

[tulip temple]

anyone have experience with autonomous business models?

im interested in trying it out, i live in Cali lmk if someone can help me out

or interested

[late atlas]

wish i was there

[dense hill]

4o keeps switching to thinking mini. How do I stop it?

[modest steeple]

how to add official chatgpt to my discord server?

[paper latch]

Anyone who knows how to contact them about the plan features?

[rapid sleet]

what's wrong with codex on vscode?

[halcyon lava]

@rapid sleet have you tried logging out in VSCode and logging back in?

mine fixed today

by itself

[rapid sleet]

It's still error😭

[rose kraken]

it is possible to integrate ChatGPT into Discord, but it requires setting up a custom Discord bot that connects to the OpenAI API.
lol I used chatGPT to find out just as you can 😛

[modest steeple]

thanks bro

[blazing moat]

AI-Dementia?
I am a Plus-User: While doing research on papers and books in a short (!) and very precise request I considered what I want to call "AI Dementia". The task was not complex - resorting a list of 13 papers in Harvard citation. But there was a severe memory loss. After the first resorting 4 authors of 13 were missing. After manual correction at the next attempt 2 books from authors with more than one contribution were missing, although I uploaded the completed list in the chat.
At my complaint GPT-5 thinking answered, that it "can't provide consistent information for more than 6-8 iterations".
My experience was even worse.
That is a no go, if you try to do research for literature.
Any ideas how to handle that problem? My prompts were clear, structured and provided all necessary information.
Anybody else with that problem?

[unreal willow]

I got this at this very moment, and hopefully this is a step in the right direction to fix many of the problems that have been reported on this channel for the past week

[untold trellis]

Hey Im new to the discord server, but I wanted to know if anyone is encountering issues with their folders? When I make a new one an older one is erased?
I'm assuming that the post above me may very well be linked

[thick hill]

Hi, I'm having trouble downloading my generated video in Sora. Is it going to be fixed?

[drifting oar]

Hi , I am using openai python sdk.
I have very high response time for gpt 5 mini model 10+s, while gpt 4.1 is super fast ~1s.

Any ideas why that might be the case?

[scarlet patrol]

I’ve had the same problem for days and it still hasn’t been fixed. I can’t download videos from either my smartphone or my PC

I cleared the data and wiped the cache, but it doesn’t help — I still get the same error when I try to download the videos I create.

[high lily]

Can someone help me with finetuning?

[sudden delta]

why is gpt doing thinking modes when i explicitly declare "instant" is what i want

can it like not do this?

[foggy tapir]

I have found that when working on projects that Chat will very often misrepresent what it is capable of doing. Say it's working on things and then take days and lots of prompting to get it to not drift and get a project complete. Also, when you ask it for instructions on how to use it, it gives incorrect info. Does anyone else have this issue and what do you do about the waisted time and effort? Especially since I pay for the privilege of being misled weekly

[harsh furnace]

Hi, I'm having issues with Sora downloads. I pay to not have watermarking and yet even though I click to donwload those versions; they're all downloading with water-marks. I cleared my cookies and whatnot but still didn't fix the issue. Is there anything else I can do that someone is aware of? Thank You Kindly

[vivid mortar]

Have you downloaded videos in the past without a watermark and this is a recent issue?

Which method are you going about saving/downloading them?

[vivid mortar]

In my experience its best to summarize your chats when you are in a good place and having a solid conversation, continuing in one chat for days on end or even hours at a time can be ineffective since each chat only has so much knowledge/context and the more you chat the more those older messages slide outside the context window.

Summarizing and keeping a "checkpoint" will help you see all the information its currently got access to and works as a nice copy-and-paste-able way to start new chats

[foggy tapir]

So copy and paste the chat into a word doc every know and then? Sometimes it is because the content in the chat is long. In this specific case I asked chat to organize and document the content ideation we did together. It told me it would take a day or so. Now through multiple chats and giving a new chat the entire conversation with asking for it to organize and document the ideas it cannot get it done. I don't understand why and it doesn't either, lol. Why did it tell me it was working on something and then admit to not working on it

[vivid mortar]

Honestly if you’re doing that sort of content planning and want it to actually edit files and create documentation I would suggest using codex in your terminal

There's a bit of a learning curve with getting everything installed but I think that may be a good solution for what you're wanting to do.

Codex CLI is a tool suite built by openAI for editing files on your local computer, you can have it create those documents and then read through the changes on them and then it can read the current state of those documents and iterate from there

[foggy tapir]

Thank you! As a realtor trying to find out how best to use AI and get ahead in it in my industry is not easy. I know what I want to use it for, but I know it's not there yet and I do not know how to use it correctly yet either. A lot of trial and error

[vivid mortar]

Totally understand! What sorts of tooling or AI use cases do you see a gap for in the real estate / realtor space?

[foggy tapir]

I think it may have to do more with what I don't know than may already be available. However, being able to do content ideation, planning and scripting is just the beginning. Then we have "items of value" like "10 DYI staging techniques to sell your home". Once you have the GPTs and agents trained on brand, logo voice and such and tied to the correct apps I would like to be able to create a video and have AI combine it with video with the correct item of value then prepare an ad or retargeting. Deploy those and track the ROI. This way I stay ahead on the marketing for leads. I know that as AI and things like Canva and API 's advance I will be able to train it to create those items of value with real time researched info and create the ads as well as use API to connect to the CRM to actually follow up as well as track the ROI. Track the stats such as conversation rates for emails and be able to write or rewrite emails so that the communication on the leads generated using these techniques are operating at the highest levels. I know that's a lot and not exactly what you ask d for and that is just the tip of the iceberg for what I know AI will be able to help me with. I just don't know when it will be ready and how I learn how to be ready to take advantage of opportunities as they become available to me for my industry

[vivid mortar]

So what you’re really aiming for is an AI system that can handle the whole pipeline from content ideation and planning, to creating ‘items of value’ like guides, then assembling them into videos/ads, deploying those campaigns, connecting with your CRM for follow-up, and all the tracking.

Is that right?

This may not be the right chat but hop over to #ai-discussions I have some ideas for you

[foggy tapir]

Yes, that is correct. That is one area I know AI systems will get to

[high lily]

gpt 4.1 nano fine tuning with bs = 16, lr-multiplier = 0.1 and epoch = 3 results were: Train loss 0.264 Valid loss 0.256 Full valid loss 0.303 . The issue is now on my eval set accuracy is 43%. Which parameters should i change for next iteration?

[sudden delta]

once again i was doing a conversation with gpt5 and out of nowhere it switches to gpt5 thinking mini and not once was i ever on auto i am ALWAYS ON INSTANT
it is stuck now and no version of my prompt can possibly revert it
It has destroyed the entire conversation and deleted basically the tone of the writing

this needs to be fixed

[high lily]

write to support

[high lily]

.

[sudden delta]

already did

no reply

no reply from support on many other tickets of the same type too

[potent halo]

Subject: Issue with Stuck Generation in SORA

Hello guys!

I encountered an issue while working with SORA.
I accidentally started a generation using a deleted image as a reference.
After that, I received a notification that the generation had failed. However, the system got stuck in the “generation in progress” state:

the process does not finish, I cannot delete the task, I cannot start any new generations.

Could you please help me with unlocking the system and restore the ability to work?..somehow?

Thank you in advance for your assistance.

its already 40 minuts like this

[drowsy sonnet]

😭😭

[drowsy hare]

me the same -.- , good to know that i am not crazy 😄 and alone with this issue. Is there a Solution? I still have this problem

[slow karma]

hi @tulip temple , I'm autonomous business man, Can you help me?

[north tangle]

Today, for me, time for video generation is very long… i think is a Sora problem

[potent halo]

well, its been more then an hour for image generation. ..and Explore page doesnt load...but they are rolling out SORA2, maybe its somehow conected

on satus.openai.com looks clear

[modest thistle]

For the last couple of days I've been getting an error message when I search my chats.
I'm a plus user on the Android app.
Is this a known thing?

[magic cairn]

Is QuizGPT down? It is not not working no matter what platform I use, it utterly refuses to use it no matter what mode I put it in.

[fossil meteor]

Hmm I need an invite code for Sora?

[versed pilot]

Hey. Infinite “GPT-5 Thinking” spinner, then Request expired on every prompt.

Troubleshooting already done
• Reinstall app
• Clear cache / data
• Disable VPN, test multiple networks
• Sign-out / sign-in, full device reboot

Environment
• iPhone 16 Pro Max
• iOS 26 (latest public release)
• ChatGPT mobile app up-to-date

Additional blocker
Opening a conversation containing an extensive Deep Research session triggers an immediate application crash.

Impact
Service has been unusable for several weeks; no accessible support or workaround identified.

[lime vigil]

Hey there I have the plus plan and it says I cant generate more than 1 image at the time... at first this was possible ofc... its the whole point of paying in the first place !

I can see from a couple of days ago an image always doing the rotate animatione like trying to create an image so... it might be a bug ? @onyx kraken any idea ?

if I enter to that image it looks like this... but this is from yesterday

How do I ask for help about this ? I dont see any option in the app... @tribal talon

[onyx kraken]

Can #1070006915414900886 to describe the bug here, and can https://help.openai.com/en/articles/6614161-how-can-i-contact-support make a bug report to tech support too if you want an OpenAI Tech support person to help you, the wait can be hours to days but that's how to get direct tech support

[lime vigil]

Thanks a lot!

[onyx kraken]

I hope you get help quickly!

[cursive lynx]

Can someone help me

[cedar wharf]

How do I get invite to new sora

[proud spire]

Hey Ya'll so I shared a code but only one code, however all 4 are gone now, is that a bug or a feature? Is it one code and regardless of the code it's good for 4 people ?

[cedar wharf]

I am a paying customer why do I still need a invite

[prime finch]

I would greatly appreciate an invite code to Sora. Thank you.

[grave hull]

I think you have to be subscribed to ChatGPT Pro to receive an invite from what I've heard unless someone you know gives it to you because I know for every person who receives an invite they receive 4 codes that they can distribute to individuals they know

[near shell]

I have Pro and still no invite!

[radiant grotto]

Same

[short slate]

So I'm stuck in a crazy authentication loop where desktop and mobile sora are acting like separate accounts. I used a sora code it seems like I have access on mobile, but on desktop it endlessly prompts me to enter a sora invite code. Now I have 2 separate usernames for sora under the same login and every requests is a 401 and 403 hahah. Can't change any profile settings or anything they all fail. Anyone else experiencing this? Can't like any videos or anything it just unselects the heart with the requests failing everytime. I think I somehow accidentally created 2 separate sora accounts under the same login and now everything is failing. Not really sure what else to do at this point there doesn't seem to be any support channels other than here?

[agile drift]

I have Pro but Android. I think only Pro users with Iphones get invites

[agile drift]

Is it four codes or one 4 time use code

[short slate]

Oh maybe that’s why my account is in this broken state I have ChatGPT for business 😭 wish I knew how to get support for this. I even tried the support GPT but it just stopped replying to me lmao I’m so cooked.

[gritty shuttle]

Anyone know why sora 2 that is suppose to be able to generate up to 20s is doing 10s, and up to 1080p is doing like 640p?

[void lagoon]

It’s probably the massive load the usage is dumping on GPU’s

I generated over 50 videos today, and noticed that at certain times of the day quality was effected

“Over” I lied, I generated exactly 50 and maxed it out

[grave hull]

How were you even able to gain access to Sora 2? Are you a Pro user? I've heard that not all Pro users have access which leads me to wonder that it's not linear in terms of Pro users still not receiving invites if that makes sense

[wild flame]

Hella guys,im intersted in getting into ai engeneering and like to ask where to start,im currently about to Turn 17 still go to scool.

[onyx kraken]

Do you already use ChatGPT?

[onyx kraken]

I ask, because in addition to asking people - ask everyone you want! Not all of us will know.

But the model likely has a lot of training about that and many related things.

Discussing your interests, what you already know, what you want to do, what you might want to learn to be better prepared to move towards your goals -as well as discussing what you find when you google, what others recommend to you - the model can help you expand into your next steps too

[plucky current]

I have the verified and unverfied roles

anyone know what's going on?

[undone finch]

What happened with the Plus plan on Sora 1? It wont let me select 720p videos anymore and only 5sec of 480p.. Anyone has the same issue? Or its just a downgrade?

[lime vigil]

so far nothing... I will add it to the bug report...

feels like a scam and then they call it a "bug" I have the same output as a non paying client... complete scam

[thorn mantle]

How are you supposed to get a code, just wait? I mean it says it’ll notify me but how long did you guys wait for yours

[ruby yew]

Hey
Quick question: does anyone here happen to have an extra Sora 2 invite code they’d be willing to share? 🙏
If not, do you know if upgrading to Pro gives direct access and if it works outside the US as well?
Thanks a lot in advance!

[dull warren]

anyone with a sora 2 invite code? Thank you in advance!

[distant cargo]

J69BSQ

speed up

[vivid mortar]

Please do not use this channel just for code requests

[manic wren]

Ah thanks for the info. Once I know summat it unobtainable (stuff hundreds a month lol) it drops out of mind lmao. Video gen is fun but not essential. I just wanted to fiddle.

Thanks dude.

🤣🤣🤣 I wish we never heard about stuff until we can just have it.

[cunning drum]

Did Someone Have Invite Code

Please Give Me

[slim vigil]

NW3HXX

[cunning drum]

Thanks Dude

[cunning drum]

Thanks

[viral glacier]

How can gen a sora code or how I find one??

[lucid hemlock]

Please i'm searching for a Sora Invite Code

[shell vortex]

Hi !! ... I saw that you’re sharing Sora invites. If you have a spare one, could you please send it to me via DM? I’ll just use it for testing/production. Thanks a lot! 🙏 DM -me mplease. is just for work, not for play.

[trail cairn]

#discuss-sora-2 and also the videos are full of watermarks

[patent bear]

How do I get invide code to Sora 2?

[onyx kraken]

I don't know that it is a bug. It COULD be, which is why I recommended bug reports it.

I'm aware Sora 2 was released yesterday, it's also possible that Sora 1 image gens to plus users is dynamic, scaled based on available resources - we've seen that in the past with ChatGPT message limits; usually for a while we got 40/3 hrs with plus; during an especially high-demand time we all were dropped to 25/3 hours for a while, then back up - eventually limits expanded to what they are now!

I bet Sora image gen limits expand too over time.

I see this:

https://help.openai.com/en/articles/10877094-creating-images-on-sora

I see it says "You can also select the number of image generations you want outputted. ChatGPT Pro users can generate up to 4 images at a time."

I don't see any statement about how many images a Plus member 'normally gets'. I understand feeling upset, I have no idea if it's a bug, if it's a dynamic use scaling, but I bet over time everyone continues to get more access.

Good luck!

[hybrid ledge]

Im searching for a code, anyone?