community-help

[fast burrow]

So I have been a heavy user of OpenAI (have spent around 2K in total since it first got released). The other day, I make an API call to the 'o1-pro' model, and it just kept running for ages, and in the end I got no output, and was charged $25 for that API call.

So I reached out to customer service to tell them, and the screenshots is how they repspond. I really think that even their 'human' customer service people are actually AI. I don't know where to go from here. Any advice appreciated.

[trail cairn]

some are bots there. with that model in api and no context cap the response can take a lot of time and result in timeout depending how/with what it was called. could also cost a lot more. depending of the endpoint the response data could still be on the server

[misty flint]

Modcheck , shotgun spammer is back

[fast burrow]

thanks for letting me know

[onyx kraken]

Hey hi. You've explained really clearly, I think I understand what you're explaining.

We're community members here, like yourself, nearly all of us; even the mods are volunteers, not OpenAI employees. The very few here who do work for OpenAI have gold names, #server-staff .

There are a few ways to reach out to OpenAI help here, and one is #1070006915414900886 .

It's probably very difficult for OpenAI to track the chats here in these channels, there's over 100k of us that may chatter about all kinds of stuff, so I wouldn't consider explaining in this channel to be 'letting OpenAI know' - but we have decent evidence they check both #1070006915414900886 and #1070006151938314300 - some even get visibly answered by gold names.

I recommend making a bug report, and in either the initial post or the follow up post, explain in detail with your screenshots as you choose. You may find that other community members may spot your report and add their own discussion about similar situations, if any has happened.

Be careful not to include any PII; you can use something like Paint and draw a filled in rectangle over any info that might be PII (community members can see these, so do protect yourself in general).

Note that there's an exact process to make a bug report, this may be your first so please do read these instructions, and if you have any trouble making the bug report, do ask for help. It's okay to reply back to this, I'll help as I can. I'm a volunteer guide - not an OpenAI employee. I can't help with the part you need OpenAI's help with. But I can help ensure you can make the bug report if you feel stuck, sometimes people say it's confusing. https://discord.com/channels/974519864045756446/1295975636061655101

[exotic ravine]

Okay table and chart making functions are now fully back online wooohoo

[rugged wing]

is there a limit to archived chats? there is this one chat that no matter how many times i click archive and accept, it won't work, and the chat will never archive. tried on desktop (safari), macos app, ios app, and nothing.

[slim marten]

Hi everyone,
I’m developing a business-critical custom AI assistant using ChatGPT Plus with long-term memory enabled. The assistant has learned a massive amount of structured information (brand logic, communication framework, design system, automation, GitHub sync, etc.).

Now I’ve hit a critical limit:

🔴 Memory is 100% full in the ChatGPT UI – and I’m just a fraction of the way through development.

I’m aware of the “Settings > Data Controls > Export Data” function, and I’ve done that.
But on top of that, following my assistant’s advice, I’ve generated a fully AI-readable YAML + Markdown archive of the entire project (ZIP format).
It includes:

• memory-index.yaml
• memory_entries.yaml
• command_map.yaml
• changelog.md
• stilona_visual_design_board_full.xlsx
• stilona_brand_summary.pdf

🧠 The goal is to preserve all knowledge with 100% precision and reusability, so nothing is lost.

---

❓Questions I need expert advice on:

1. What’s the safe strategy to clean up memory without losing critical AI learning?
2. Are there any hidden options or workarounds to increase memory capacity (aside from switching to GPT Enterprise, which I can’t afford)?
3. Can I store and reload my memory externally using a RAG architecture (e.g. Google Drive + ChromaDB)?
4. Is it possible to simulate long-term memory through structured YAML documents that are regularly reuploaded?
5. Any best practices you recommend for handling large-scale AI projects in ChatGPT Plus?

---

I’m looking for serious, experienced input – either from OpenAI team members or power users who have solved similar problems.
Even just pointing me to the right forum/channel/resource would be incredibly helpful.

Thanks in advance for any guidance or experience you can share!

[fast burrow]

Thanks very much! This was really helpful!

[azure halo]

I saw that you also asked the same question in the developer community. I have also faced the same problem several times, as I use ChatGPT very intensively both professionally and privately.
Unfortunately, ChatGPT also has a habit of saving reminders from time to time that are irrelevant to me. Some of the information has also been stored twice by the AI.

This makes it even more difficult and it is indeed difficult to decide what can be deleted. More flexibility would be really helpful here.

[round cedar]

hey everyone. For the past two months i am being billed 23.50$ instead of 20$. Does anyone know why?

[trail cairn]

taxes, but you can check your invoice

[graceful lark]

Is anyone else unable to interact with Operator's "Take Control", or barely at all? Mine lags so bad it often takes 30 seconds for it to receive one input

[slim marten]

Unfortunately, no one is responding who would be at a more advanced level and have already solved such a problem. But even the OpenAI developers are staying quiet...

[sweet pulsar]

why is this happening when i ask gpt o4 mini a maths problem?? its response gets deleted and its VERY annoying

[onyx kraken]

What's the prompt you're inputting?

[sweet pulsar]

an image of a sparx maths question

forgot which one

[graceful lark]

Has anyone found the API equivalent of image prompting in o3?
In o3 I can do sophisticated, masked editing with a single prompt but the API is struggling.

[onyx kraken]

Without the prompt chain, it's really hard to try and find out what might cause the refusal.

One option you have to report it to OpenAI, won't help you get an answer why like you asked for, but if it's a bug can help them find and fix it. https://openai.com/form/chat-model-feedback/ Explain the issue, paste the chat link (they don't need a share to be able to look into it), and they'll explore a bug or whatever training they need to do if there's a problem.

[thorny crown]

??

[sweet pulsar]

👍

[acoustic turret]

hi, i got isue , not able to create video or image. error.: something went wrong,please try again later. pls fix it asap.

my work is impacted.

[ebon storm]

why getting this error two times in a row and also why chatgpt is rendering slow.

[vast forge]

#status

[gaunt jacinth]

GPT 4.1 and GPT 4.1-nano are extremely slow. Has anyone experienced this?

[swift heron]

Hey guys, ChatGPT and it's services are slow and it's down
So, as of a minute, please check #status , thanks.

I also found these posts on Threads.

[jolly kelp]

is anyone getting this: Too many concurrent requests

[static talon]

yup chatgpt is down atm

[jolly kelp]

how come they haven't fixed the issue yet 🙄

[chrome orbit]

Getting the same error. Plus unable to edit custom GPTs....gotta wait i guess

[queen oasis]

Hello, is there someone knowledgeable in gpu and stt that might be able to help ?

[elder turret]

Error in message stream

Retry

i have the same issue

[final bridge]

I've been getting this message when using chat "I wasn’t able to generate that image because this request currently violates our content policies. This may be a false flag, and I understand it can be frustrating" . I had asked for a quote than to create an image with the quote. this was fine before but been getting error for like a week now

[willow coral]

why does data analysis never work:
use data analysis: oh my god let me write all this code error
use reasoning: the thing has code in its reasoning chain and sucks

[dense helm]

why where a bunch of tabs removed? its way more boring to go to the my custom gpt page now...

it was still here until like 30 minutes ago

[prime cairn]

Hello, can You help regarding making self pictures into LinkedIn profile pictures is frustating since its making a whole New person, someone recommend a strategy prompt?

[dusty gulch]

My GPT was taken down which I think was done in error, but I don't have an appeal button present when I go to edit my GPT. Does anyone have the same problem or know how to fix it? Thanks in advance!

[eternal monolith]

guys, does o3pro is using "reference chat history" for you? Because for me it can only access the classic memory, not the new one.

[fringe hare]

Does anyone know how long it usually takes for sales to get back to you after you contact them?

[waxen flame]

Can anyone help me, why is my desktop chat not syncing with my mobile chat? I have finished generating a chat on mobile, but on my desktop it says something went wrong

[sturdy jewel]

have you tried closing and reopening the desktop app?

[waxen flame]

I've done it, but it's still the same

[sturdy jewel]

maybe try uninstalling it… You might have to look up if it’s storing cached data somewhere that you would also wanna remove

[waxen flame]

mobile and desktop?

[sturdy jewel]

Just the desktop one if it’s the one messing up

[waxen flame]

wait i will try it

thank you

[waxen flame]

still same

[sturdy jewel]

what is your desktop OS?

[waxen flame]

Windows 11

I tried opening it in a browser and still error

[sturdy jewel]

that's weird, is it a really long chat?

[waxen flame]

humm yes

on my mobile it's done, and it appears, but on my desktop it doesn't appear

[sturdy jewel]

really long chats can just mess up sometimes, you might have to just copy the relevant context into a file and start a new chat

[waxen flame]

humm okay.
thank you for helping me.

[sturdy jewel]

np have a good one

[eternal monolith]

guys, is o3pro using "reference chat history" for you? Because for me it can only access the classic memory, not the infinite one.

[sturdy jewel]

seems like it works for chat history

[eternal monolith]

oh, but thats not o3pro mode

[sturdy jewel]

yeah I can't help you there 😄

I just have regular o3

[eternal monolith]

thanks for trying tho

[chrome orbit]

hi guyssss. Hows everyone doing? Been trying to make a custom GPT. But not able to UPLOAD files into it. The button is simply not working. Would love help from anyone who knows about custom gpts

[onyx birch]

Hi,
I'm wondering if, and when, the Linear connector will become available to the EEA.
https://help.openai.com/en/articles/11487775-connectors-in-chatgpt#h_25ad24932c

[wet zenith]

where and how do i share my website on here if thats allowed at all?

[onyx kraken]

Nearly always, non-OpenAI website links are blocked here. Most of the reasons why people want to share websites often conflict with #server-rules , like we can't solicit or promote.

There are a few channels that allow some urls, but not many.

If you have a reason to want to share your website that you feel is deeply within the allowed rules, perhaps ask modmail for assistance.

[wet zenith]

It's a website about my research and interactions with my AI

[fresh flume]

Is there a feedback portal?

[timber shadow]

Are you looking for a place for general ChatGPT feedback? If so, the support chat at help.openai.com is a good place to give feedback. We also have #1070006151938314300 in this server if you want to see if anyone else here has shared similar ideas!

[true aspen]

Hi

I would like to know approximately how long it takes for Open IA to respond to us for a serious project in Modmail and also in an email to partnership@openia

[mint geode]

has anyone seen this problem before where the teams account has been deleted, and user is unable to create a personal workspace?
I have tried clicking the "Create Personal Workspace" so many times but nothing happens.

[trail cairn]

modmail is only for discord issues

[true aspen]

Ah sorry my bad

But you You know, a little ready to have a response on partenership@openia I think it is reserved for commercial collaboration so I should have sent it to the classic email I think

[trail cairn]

i dont know where you got that email but ive never heard of it and openai uses only/mainly web contact forms so it might be outdated or unmonitored

[onyx kraken]

I have never been teams so I only know what the helps say.

Here's a few that you may want to look at, instructions how to switch, in case that's different from what you're doing:

https://help.openai.com/en/articles/8798594-what-is-a-workspace-how-do-i-access-my-chatgpt-team-workspace

https://help.openai.com/en/articles/8798618-what-happens-if-a-workspace-has-been-deactivated

I don't know if leaving the deactivated team workspace might help:

https://help.openai.com/en/articles/9123978-how-can-i-leave-a-deactivated-workspace

Here's info about if you merged your personal workspace with your team workspace previously:

https://help.openai.com/en/articles/8801890-can-i-migrate-or-merge-my-chatgpt-free-or-plus-workspace-over-to-my-chatgpt-team-workspace

The bottom right of those pages, there's an icon you can click to get help. Initially you talk to a bot, select messages , send us a message, get started, then click through the choices until you can report a bug. That'll allow you to explain the problem, at first to the bot, and be played in a queue for human help.

Sometimes an account may need assistance on their end, maybe it isn't always 'oh, click here', so I would ask tech support for help if the help page instructions don't solve it.

[onyx kraken]

https://openai.com/form/chat-model-feedback/, you can privately give them the chat urls that show the problem through that as well

[raw bridge]

O3 pro is really not that smart

[daring flare]

heyy, has anyone worked with OpenAI assitant, i have a few doubts around it

[slender sphinx]

Hey quick question, does anyone know if the OpenAI team plan has a per user limit on image generation?

[frank socket]

help, I paid the plus subscription again from my iphone and I get that I pay to another account and from there to start the subscription, what do I do?

[rustic marlin]

contact the support at https://help.openai.com and explain the situation, you will be refunded

[slender sphinx]

lmao I wonder if anyone falls for random images...

[lofty panther]

Hello, I was searching for a way to configure VAT-less invoice and the payment window ran out. My workspace is deactivated and when i log in there it says loading the settings failed. When I login I think it logs into the personal account, I can't see any payment options nothing. How do i get acces to the admin workspace so I can pay modify invoice settings. I must say it is weird how complicated this is. (I upgraded to Teams from my personal workspace, and I have a feeling it's a bug)

[hoary shell]

Hi everyone, I'm currently working with the Model Context Protocol (MCP) following the guide from this link https://platform.openai.com/docs/mcp . I managed to set up the server and tested it using Postman to check if I can establish a connection and make requests. Everything works fine there. Then, I exposed the server using ngrok and connected it to ChatGPT through a connector. When I added the connector, I could see that ChatGPT successfully connected to my server, made 3 requests, and then closed the connection. The connector appears correctly in the interface, but when I enable the deep research with this connector, ChatGPT doesn't seem to use it.
Has anyone encountered a similar issue or has suggestions on how I could troubleshoot this?

[median violet]

hey guys
does anyone know why deep research is not detecting my custom connector sources I am sending them just like the docs said and still 0 sources does anyone know the reason for this ?

[warm python]

my images keep getting flagged in sora when all I'm trying to do is remove my very own logo that i designed from one of my mockups

any way to fix this so that I can continue fleshing out ideas?

[thorn condor]

Hey guys, I just upgraded my Plus account to a Team plan, thinking I would be able to keep using the same account. Now I have two accounts— is this supposed to happen?

[golden ruin]

hey! this may be helpful https://help.openai.com/en/articles/8801890-can-i-migrate-or-merge-my-chatgpt-free-or-plus-workspace-over-to-my-chatgpt-team-workspace

[thorn condor]

Thank you @golden ruin . You are awsome!!

[mighty ibex]

Hi everyone! So, there is a problem with my ChatGPT. I already tried asking IT whats wrong, but it just kept repeating the same thing.

The problem: Memory.

I like working with quite complex things, like speculative biology. It requires large amounts of memory storage to go on. And this wasnt a problem, like at all. GPT could remember like 4 billion years of planetary history, without any summaries, and work just fine. Now, it seems like a drastic amount of the memory storage was deleted.

It can only remember from 0.5 to 1.5 billion years of planetary history, even WITH highly summarized text. And although this may seem like not a big problem, but it really is. As ChatGPT, expecially to my projects, has to remember a lot of stuff, like planetary parameters, body plans, naming systems, metabolisms... Without remembering that, it just makes up data. Like when i ask it "What is the most popular metabolism on X, at time Y?", it responds with something totally obscure, often not even close (Note: This often happens in the SAME chat as the project is).

The memory drop has happened like 3 or so weeks ago for me, and from that nothing has changed. Does anyone know how could I make the memory work normal?

(also, the thing it was repeating was: "Recently ChatGPT has updated its memory storage capabilities." or something similar, even though the opposite was happening to me)

[trail cairn]

which plan do you have? plus?

[mighty ibex]

well actually no, i have the free version. And ive allways had it

[trail cairn]

free users only have 8k context window so you can get the goldfish memory effect fast with lots of text. you might not always notice it because it depends a lot per thread

[mighty ibex]

but still, i dont quite understand why in my previous projects everything was normal, and then suddenly this memory stuff appeared. Plus, as i said my older projects for sure had a lot more words, as when doing them in Word (font size 12) , they took up as much as 40-50 pages, but now it is more like 10-15 pages

[trail cairn]

like i said it can depend a lot on how the conversation proceeds in other terms than just original info. it just sounds the work you do can be near the limits of free plan. you can ask this from chatgpt with follow-up questions: what does this mean "free users only have 8k context window so you can get the goldfish memory effect fast with lots of text. you might not always notice it because it depends a lot per thread"

also chatgpt has real projects feature but with free i guess you just mean a topic with it. https://help.openai.com/en/articles/10169521-using-projects-in-chatgpt

[valid flax]

Why is he so idiot lately?

[frank light]

dalle in chatgpt chat session not working is there anyone facing the same issuees

[white scaffold]

Hi, I need help to regain access to GPT Models.
Recently I got suspicious activity banner. I immediately changed my password and log out from all devices. But still the same banner is appearing when I do chat

[timber shadow]

Hello, do you see this message when you try to use the ChatGPT app, or chatgpt.com, or both?

[white scaffold]

I am only using it to web. chatgpt.com

[timber shadow]

Got it. Does it happen on every device/in all browsers? I think typically a message like this might be caused by connection to a VPN, some browser extension(s), etc., so experimenting with disabling those might be a good test, if you haven't already

Otherwise, OpenAI support at help.openai.com may be able to point you in the right direction

[white scaffold]

I am not using any VPN at the moment. Also I have cleared site data, cookies and disabled extensions.

[potent ivy]

Can I make a post in the prompt engineering chat requesting someone to review my beginner prompt & provide feedback? It'd include a URL to GitHub.

[signal phoenix]

Anyone experiencing unreadable math out put strictly while using the Deep Research feature ?

[gaunt grove]

im having a problem with chat gpt. if the chat goes on for too long (too many messages), the UI starts lagging a lot, theres a significant delay when chat gpt displays text or when i type text in the chat. chat gpt says this is because the UI is rendering the entire chat log even if the browser can only display a small portion of the log.

im trying to use the workaround where you tell chatgpt to remember the messages and resume in a new chat window, but chat gpt keeps forgetting things it should have remembered and i have to spend time reminding it. E.G. if i say "i already tried X", tell chatgpt to remember everything, and resume the conversation in a new chat, there is a high chance that chatgpt will forget i said "i already tried X" and recommend i try X again.

does anyone know if there is a better solution to this?

[white scaffold]

Hi, I need help to regain access to GPT Models.
Recently I got suspicious activity banner. I immediately changed my password and log out from all devices. But still the same banner is appearing when I do chat.

Steps I had taken:
Changed Password
Enabled 2FA
Cleared Cookies/Site Data on Browser

Tried changing intenet connection

Despite all these steps I am unable to regain access????

[brisk pewter]

HI all , i am unable to generate images thru dall e . is there something wrong with interface ?

[maiden elbow]

Hi, I subscribed to ChatGPT Plus on June 13 at 5:00 PM KST via KakaoPay, but the subscription is not active on my account. Can you help me restore the subscription?

[zenith solstice]

hey folks, quick question — I can’t seem to access the Scheduled Tasks feature in ChatGPT.

I tried typing stuff like “remind me every day at 9am to do X” right in the chat, but nothing happened. No setup, no task created.

weird thing is, I do have an old task ("AI summary reminder") that still shows up under Notifications, and I can edit it. so the feature was definitely there before.

also, the tasks option doesn’t show up under my profile icon or anywhere else in the UI.

I’ve already tried GPT-4o, switched to English, tested on desktop and mobile… still nothing.

anyone else run into this or know how to get it working again?

[timber shadow]

Hey! Currently you can create new tasks with o3 and o4-mini: https://help.openai.com/en/articles/10291617-scheduled-tasks-in-chatgpt

[zenith solstice]

cool sorted thank you so much!

[timber shadow]

Hello, did you subscribe on chatgpt.com, or in the ChatGPT app (iOS, Android)?

[maiden elbow]

I subscribed from Android

(App)

[timber shadow]

Gotcha. One thing to check if you haven't already: make sure you are using the official ChatGPT app from OpenAI, you can find the link for Android here: https://openai.com/chatgpt/download/ If that's all good, you might check your Google Play invoices to confirm that the Plus charge went through as expected, since you paid through the app. If both are good, then hopefully this is just a temporary bug/delay you're running into, and hopefully it resolves itself soon. You can reach out to OpenAI support at help.openai.com too -- if nothing resolves then you'll definitely need help from someone at OpenAI

[tender badge]

Too weird
Weir

[harsh gull]

i am looking for help in writing a prompt for my Pensioners club. the prompt i would like is as follows:
Given a diagnosis by a medical consultant, check against a list of medical support provided by a medical insurance company, giving the exact Code | Diagnosis the injury falls under, for example, as follows:
| A6030 | Transection of peripheral nerve for neuroma. Any help will be greatly appreciated. Also suggest the OpenAI model to use. thanks in advance

[vale trench]

What are you looking to get as output? Do you know the exact format of the list from the insurance company? I don't fully understand what your goal is.

[forest cipher]

the AI is unable to create new canvases and claims that openAI disabled this feature. when i open existing canvases i cannot edit the text (through the mini prompt line). can anyone help me or give me more info on this? EDIT: I am on Brave browser.

[gaunt grove]

im having a problem with chat gpt. if the chat goes on for too long (too many messages), the UI starts lagging a lot, theres a significant delay when chat gpt displays text or when i type text in the chat. chat gpt says this is because the UI is rendering the entire chat log even if the browser can only display a small portion of the log.

im trying to use the workaround where you tell chatgpt to remember the messages and resume in a new chat window, but chat gpt keeps forgetting things it should have remembered and i have to spend time reminding it. E.G. if i say "i already tried X", tell chatgpt to remember everything, and resume the conversation in a new chat, there is a high chance that chatgpt will forget i said "i already tried X" and recommend i try X again.

does anyone know if there is a better solution to this?

[somber hill]

Have you recently have it that your last parts in chat simply disappear, are no longer there?

Habt ihr das neuerdings auch so, dass eure letzten Teile im Chat einfach verschwinden, nicht mehr da sind?

[somber hill]

Oh, I hadn't read that exactly. It will be.

Oh, das hatte ich gar nicht genau gelesen. Das wird es sein.

If you have the last message read out, he also reads the penultimate one. So both are a message.

Wenn man sich die letzte Nachricht dann vorlesen lässt, liest er auch die Vorletzte mit. So sind beide dann eine Nachricht.

[gaunt grove]

what?

[somber hill]

Go to "read a message". Then the penultimate and last message will be read to you with the text to Speech model.

Geh mal auf "Nachricht vorlesen". Dann wird dir die Vorletzte und letzte Nachricht vorgelesen, mit dem Text to Speech Modell.

[gaunt grove]

this is not helping me with my problem

[somber hill]

Yes, We have the same problem. Your own message is just gone.

Ja, wir haben das selbe Problem. Die eigene Nachricht ist einfach weg.

[heavy patio]

Personally speaking, a lot of my canvas issues are usually browser related. Even when using a standard browser, if I get out of memory and have to refresh or comments will stick on the screen and require me to refresh the session. It's an incredibly new technology, so I try to bring patience.

[somber hill]

I am connected via the Chatgpt Desktop program. I never use Canvas that the small model does not want to process and I don't want to limit our chats like this. Canvas is probably very important to others.

Ich bin über das ChatGPT Desktop Programm verbunden. Canvas nutze ich nie, das das kleine Modell das nicht verarbeiten will und ich unsere Chats so nicht limitieren will. Anderen ist Canvas da wohl durchaus wichtig.

That with the two messages is also the case in the Android app.

Das mit den zwei Nachrichten ist auch in der Android App so.

[heavy patio]

I've had personally have super long chats a lot less since I moved a lot of my conversations into "topic groups" that are now 'projects', and started making new conversations more often. I find just editing a core set of "knowledge" on the topic in files (documents we produced of things we want to remember), and instructions to follow.

The extra step of organization really makes the AI much more efficient and grounded in the behavior I want to see for that "topic" (project).

As of a few days ago project memory got better, also!

[gaunt grove]

the problem is that i frequently have to start a new chat because the chat drags on too long and it starts lagging. so you are dealing with it by uploading files to make chatgpt remember stuff every time you start a new chat?

[signal phoenix]

Anyone having problems with Deep Research and math notation displaying in an unreadable way ? its in tex or latex code rather than the normal mathmatical display. Only happens in deep research

[lime flint]

@exotic fiber

[exotic fiber]

Thank you

[lime flint]

❤️

[exotic fiber]

This is the way.

[lime flint]

@exotic fiber they are back -_-

[harsh gull]

What I would like is the ability to compare the consultant's diagnosis to the list of insurance policy coverage and draft a letter refuting the reason why the request was denied. is this clear, sir?

[vale trench]

I can try to create such a prompt for you. Would you mind sending me the initial task, this last message of yours, and potentially a bit of example data via DM? I can try and make a prompt for you, but since I'd need you to check out the output I think it'd be a good idea if you DM me anyway.

[harsh gull]

than you so much, i will.

[lime flint]

How can I fix this?

[normal pendant]

Are you using a vpn?

[river geode]

You're supposed to be able to have 10 active tasks in ChatGPT scheduled tasks, but the AI model has been telling me that the limit has been reached and that it can't make more when I only have 8 or 9 tasks. What's up with that?

[round ember]

Not sure if this is the right channel, but I’ve been struggling to animate a photo in Sora — I must have tried like 30 times 😅
I’ll share the image below.
Could someone help me and write a good prompt for it? I’d really appreciate it!

[somber hill]

Yes, my translator translated from German into English to express me fully, I would have to be a native speaker. What I am not. So I express myself in German fully and translate the translator. Albeit more badly than right. The Google translator is not good.

Ja, mein Übersetzer übersetzt vom Deutschen ins Englische Um mich vollumfänglich auszudrücken, müsste ich Muttersprachler sein. Was ich nun mal nicht bin. So drücke ich mich auf Deutsch vollumfänglich aus und der Übersetzer übersetzt. Wenn auch mehr schlecht als recht. Der Google Übersetzer ist nicht gut.

[obsidian stone]

Not sure if this the right channel, but chatgpt seems to not be able to read the project files it has. Refuses and said it is to protect my privacy? I'm confused.

[golden ruin]

hmm have you tried in different chat?

[obsidian stone]

Yes I have. I even created a new project and got a similar answer

[obsidian stone]

I even tried a text file with "hello world" in it.

[rugged wing]

I think I’ve encountered something like this. I’m gonna look into this. I’ll give you an answer later on. I forgot what it was that fixed it.

[obsidian stone]

Thank you!

[rugged wing]

Hey real quick did you check your settings on your ChatGPT?

[obsidian stone]

The only thing I turned off is using my info for training (company requirement)

[rugged wing]

Can you upload PDF? Or do you get the same issue?

[obsidian stone]

I'll check. I even checked if it was something I did.

It’s not about policy or caveats you mentioned; it's a built-in technical limitation: I don’t have file‑reading capabilities in uploaded project workspaces.

[rugged wing]

Are you using the API?

[obsidian stone]

Joy

[obsidian stone]

The UI. We are having a splendid argument. 😂

[rugged wing]

We’re on two different pages my friend

[obsidian stone]

i was on my phone. let me go to my desktop

[rugged wing]

You’re still having the same issue?

[rugged wing]

And you can’t add nothing to it?

[obsidian stone]

one sec. finding a pdf on my computer i can use

[obsidian stone]

well. this is a new one

[rugged wing]

Oh ya

Looks like you got the case of I don’t know what the hell is wrong with your ChatGPT

[obsidian stone]

yeah, the server error makes me think there's something going on i just need to go for a walk and wait it out

[rugged wing]

You know, now that you mention it, I’ve had some issues uploading some images the last few days and I believe they ended with the same error you showed me above with the same ORD errors

[obsidian stone]

let me look at the dev tools console log..

welp.. nevermind

nothing

oh.. that uploaded...

[rugged wing]

Check your settings

In your memory

[obsidian stone]

waaaat... that worked

not sure where that one is

[rugged wing]

Damn, you got a whole bunch of other stuff on yours that I don’t have

[rugged wing]

You got a quick customize, ChatGPT

[rugged wing]

Yeah, looks like everything’s on for you

If you don’t use any of that stuff, just keep it off. I try not to mess with any of it. I just keep it the way it is.

Well, first try it with it on see if it helps

[obsidian stone]

wuuut... i renamed it from "Policy Template.md" to "Policy Template" and ... it worked

i'm going to try .txt

uhhhhggh i might just trash everything and start over

how many hours i wasted on this... unchecking that seems to have fixed it

[rugged wing]

Cool man awesome now you know for the future, the diagnosis haha

[obsidian stone]

let's see if this works

[ionic torrent]

Did they completely remove dall-e 3 from chatgpt.com?? the gpt won't generate any images?

[bleak goblet]

Is Deep Research updated to the latest version of o3?

[runic steeple]

So ig its a ios issue. Mhm

[sterile anchor]

learning with ia is not recommended

[runic steeple]

@timber shadow

Weird. Msg didnt sent jn…

[timber shadow]

Hmmm does o3-pro appear when you go to the web version at chatgpt.com?

[vivid mortar]

Hey there! You can definitely still generate images with ChatGPT, please ensure that you have the 4o model selected

[ionic torrent]

I guess I assumed the dall-e GPT would have the correct model defaulted to it. Thanks

[torn raptor]

Hi, I run into an issue while using gemini in OpenAi Agent SDK. When I have triage_agent to call handoff agents, it failed with following error.

File "/Users/.pyenv/versions/3.12.0/lib/python3.12/site-packages/agents/models/openai_responses.py", line 256, in _fetch_response
return await self._client.responses.create(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/.pyenv/versions/3.12.0/lib/python3.12/site-packages/openai/resources/responses/responses.py", line 1920, in create
return await self._post(
^^^^^^^^^^^^^^^^^
File "/Users/.pyenv/versions/3.12.0/lib/python3.12/site-packages/openai/_base_client.py", line 1762, in post
return await self.request(cast_to, opts, stream=stream, stream_cls=stream_cls)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/.pyenv/versions/3.12.0/lib/python3.12/site-packages/openai/_base_client.py", line 1562, in request
raise self._make_status_error_from_response(err.response) from None
openai.BadRequestError: Error code: 400 - {'error': {'message': "Invalid 'input[1].id': 'fake_id'. Expected an ID that begins with 'fc'.", 'type': 'invalid_request_error', 'param': 'input[1].id', 'code': 'invalid_value'}}
Unclosed client session
client_session: <aiohttp.client.ClientSession object at 0x1266ab950>
Unclosed connector
connections: ['deque([(<aiohttp.client_proto.ResponseHandler object at 0x1266bcef0>, 910490.300728416)])']
connector: <aiohttp.connector.TCPConnector object at 0x1266aa7e0>

pydantic 2.11.7
pydantic_core 2.33.2
pydantic-settings 2.9.1

openai 1.88.0
openai-agents 0.0.18

gemini_model="litellm/gemini/gemini-2.5-flash-preview-04-17"

[strong marten]

Where do you submit your llms.txt? Is there more you can do after adding it to robots.txt?

Google, Bing and other search engines allow submitting sitemap.xml files. Is there something similar for llms.txt at OpenAl / ChatGPT ... (or Bing Copilot, Gemini, Anthropic Claude)?

[tight vault]

Yo, my ChatGPT history all went

And I didn’t delete anything?

[trail cairn]

try with incognito mode

[tight vault]

Will try and gbty

[tight vault]

Didn’t work, but it remembers everything, so I guess all is not lost

[ocean geode]

is there features missing iuin the chatgpt windows desktop version?
like screenshare and camera?

[honest ledge]

Anyone with any insight if/when we'll be able to choose a model within Codex? I'm really curious how o3-pro would behave there.

[narrow mountain]

Since today, o3 for has switched to a loading bar instead of just "Thinking..." and responses it used to take a couple minutes to think about now take less than 10s and are considerably worse. I don't see anyone reporting this though, can this be a me issue?

[sterile moon]

is anyone else struggling to use sora to turn animated images into video? i keep getting guideline violation? love any help

[steep mesa]

anyone else having an issue with gpt right now? i keep getting "message too long" which usually means i need to start a new chat, tried that then got "error in message stream"

[lavish basin]

same here

[sly kestrel]

Same

I thought I did something wrong

Guess we'll have to wait until they fix it

Of course, this happens just as I really need to use it...

[tender badge]

Be sure.
There's no words that could be seen as a copyright

[thorny rivet]

Hi Everyone,

I urgently need assistance. My workspace has disappeared, and I currently have no access to anything. I’m also unable to make a payment because my credit card was declined, which may be contributing to the issue.

I've already contacted [e-mail], but I really need urgent help resolving this. Please let me know what steps I can take to restore access.

[sturdy jewel]

no one here can help you, go here and click on the little chat thing https://help.openai.com/en/

[thorny rivet]

[gaunt grove]

Lots of people are creating images right now, so this might take a bit. We'll notify you when your image is ready. is anyone actually able to generate images? i just get this message, then nothing forever

[rugged wing]

I’m assuming you’re using ChatGPT?

[gaunt grove]

yes

[rugged wing]

OK, so this has been kind of an ongoing issue and I’ll try to explain it the best I can so they recently updated the pipeline from the Legacy model, which was Dalle 3 degenerate images into the new image generator pipeline which could also be connected through sora

Can you check if your ChatGPT has this?

[gaunt grove]

looks more like this to me on desktop

[rugged wing]

My apologies that was a typo. What I meant to say is not search chats. I meant to says gpts

And then you should be able to see where it says dall-e

Then press star chat, and it should open you up to the Dalle gpt

[gaunt grove]

this ui is completely different from what i see on chatgpt.com

[rugged wing]

I understand we’re just diagnosing right now

What we’re trying to do is see if you’re able to generate images in general by generating through here. I could tell you if the issue with this or if it’s something else

[gaunt grove]

where is that site you are using? actually...chatgpt finally generated an image so this may not be necessary

[rugged wing]

It’s not a site it comes with your ChatGPT

[gaunt grove]

but i do not see any way to select gpts or dall-e on the chatgpt site?

does that require a paid subscription?

[rugged wing]

Oh, that’s a good question. I’m not sure. I know that I have it.

But yeah, that could be a reason why

[rugged wing]

Do you not have a subscription? I’m not getting a reply and I’m gonna take this as a good sign that it worked and that you’re able to generate in your problem was fixed.. lol

[gaunt grove]

i dont have a subscription

[rugged wing]

I think it’s gonna be kinda hard to generate without an account because I think there’s a limit how much you could generate with a free account

[gaunt grove]

yea i just hit the limit...and chatgpt does not seem to understand my instructions

[tropic belfry]

How do I remove my self from the current theme notifications?

[timber shadow]

At the top of the channel list you should see a section Channels & Roles

In the roles bit of that (section "I want to..."), you should find the section to remove that role

[tropic belfry]

Thank you

It's kinda weird. I can do things that are complex, then a simple thing like a role change I pop a circuit.

[timber shadow]

😁 No worries at all! I think in other servers it can be a different process so I don't blame ya

[tropic belfry]

Wait till my adhd kicks in.

[glacial vessel]

Hi, I get this message every time I generate an image on Sora. I have to refresh the page, and even though the image gets generated, it’s annoying to reload every time. Can someone help me fix this error, please?

[tired ridge]

hey um there is a problem with my chat gpt on my computer

it says my session has expired

and then immediately says

content failed to load

and when i press try again

nothing happens

please i need help

[gaunt grove]

so uh, can anyone suggest a free AI art generator that will let me upload a reference image, then generate an image in that style but making some changes such as changing color and other details? chatgpt's free image generation is very limited unfortunately

[tired ridge]

ye i got one

craiyon

[tired ridge]

or playground AI

[gaunt grove]

i remember trying craiyon before, but it seemed very confused and just kept generating similar images while ignoring my prompts

for example, i wanted to generate a character in a blue outfit, based on the reference image's style...and it kept generating images similar to the reference image, without making any attempt to change their color...

[wraith ingot]

I want to cancel my subscription but I get this error every time I have tried everything including support but nothing they say 2 to 3 days we will send you a message

[spiral lava]

Can someone guide me from scratch. I'm a btech cse aiml student (1st sem is going to start from August onwards). I'm a complete beginner.

[wintry flare]

Hello, I have purchase a subscription with OpenAI - ChatGPT.. I am trying to create designs and chat wont follow my commands. I have a specific mascot and I want to make different variations and specific designs (with specific text and specific additions). However, chat wont follow my instructions and will keep producing the same picture. I am using ChatGPT 4o any suggestions ?

[severe burrow]

INVITATION for FREE TRIAL: hi, I'm a beginner. On OpenAI site they write: "Eligible ChatGPT Plus users can generate a limited number of unique referral codes to offer free trials of the Plus plan." Could someone send me an invitation? Thanks

[faint notch]

Hello! I was wondering, is there an ai that can help me make notes for vids with some pics

[clear adder]

hi, anyone knows openaisecurity email?

[short hawk]

I've paid for ChatGPT plus, but can't access it

I previously had a subscription on this account through the website

Now, I reactived my subscription through my iPhone app, and I can't access it

Is anyone else experiencing this?

[gritty laurel]

Hello, I want to contribute OpenAI. I am IT developer with deep background knowledge about chemistry. Of course, I am really surprised by AI's ability, but When I am using it for chemistry, I found some serious problem.

[trail cairn]

https://openai.com/form/chat-model-feedback/

[rugged wing]

When you sign up or log in using “Sign in with Apple”, even if the email address is the same as the one you might manually type in for an email/password login, Apple creates a separate authentication pathway. So you end up with two different ChatGPT accounts:

One account tied to Apple login (OAuth token-based).
Another account tied to email/password (even if it’s the same email address).

This means:
If you purchased ChatGPT Plus on the Apple login account, it will only show up there.
Logging in using just the email + password won’t access the same account, and therefore won’t show your subscription.

I remember this happening to me

[manic egret]

how long yawl think it would take someone to build this like if i only use GPT

[gaunt grove]

chatgpt seems to be stuck trying to generate images again : ```Processing image

Lots of people are creating images right now, so this might take a bit. We'll notify you when your image is ready.```

and the weirdest thing is that if i refresh the page, chatgpt seems to cancel the image generation, but it uses up my quota for free image generations...

what am i supposed to do when it gets stuck?

huh, it seems like i need to open chatgpt in a new tab to see the generated image

[hard hollow]

Is anyone here in the Cybersecurity world and a dev? i have questions, possible collab if interested

[manic egret]

like what form of cybersec, crp? or like basic sha256 or netsec?

[hard hollow]

to keep it general along the lines of either pentesting or bug bounty, but thats more for intrest allignment than necessity for the developement.

[manic egret]

oh.. red teaming, got it. why not just say that.

not my forte, sorry

[hard hollow]

like i said, not necessary, im just wanting/trying to dev a fairly complex application with no dev experience lol

[manic egret]

im a dev. i deal with iso compliant data security, i dont deal with red teaming, but im pretty well versed.

not 15408 tho. just to be clear.

[hard hollow]

i guess my main question if continuing to try and solo dev this with AI backing, any advice on platforms or tools,including ai, i can use to efficiently bring my idea from an idea to an full application or at least an mvp

[manic egret]

thats pretty vague bro, can you like elaborate

[hard hollow]

lol basically im just trying to figure out if continuing to have GPT 4o as my dev guide is more efficient than using potential no/low code app builders. with a kali or similar linux os for the desktop gui application

[manic egret]

are you using the model window or api?

[hard hollow]

i dont think im the only one trying to make something like this obviously, but im pretty sure there isnt anything currently on market with the capability im seeking, so im trying to keep from spilling potential IP to the masses lol

im currently using 4o from plus, to dev guide and as it sits have the api running in app for functional use not dev

[manic egret]

sorry, what im asking is - how are you getting said dev guidance, via the model window, or* via api

[hard hollow]

ah through the window for dev guidance

[manic egret]

like are you limited to the 128k token window, context breaks, hard reset on local ram, cache purges? or did you outsource that locally via api by building around GPT, my understanding based on your previous reply is the former, in that case, your gaining context via project files or uploading a doc? did you develop protocols for that data in terms of industry standard development - these are relevant because it establishs the scope of the potential of your dev speed.

i.e. if you create ai understable protocols that your model follows through the rehydration of that context, sure, its easy. if you didnt, it will fail because you have to rebuild that everytime. 2000 loc processes depending on the nature of the code can require a level of hydration you wouldnt be able to inherently produce using the windos.

that prolly sounded like textbooky - let me reword that - using GPT as a dev is great if you create a replicatable structured enviorment, you dont need the api perse, but it mnakes it way easier bro, using the basic model window can work it just requires more work to make it work, so if your cool with that yea, otherwise your blackboxing your code from someone elses agent pipeline - no beuno bro bro

i would suggest, since you arent doing it, and care about MVP stage, to ask your model to teach you how to use VS code to produce a local chat bot, that you can rehydrrate with your chatGPT chats.

[hard hollow]

so far, as im still in early dev without a whole lot data in place, to keep the window convo consistant, i upload a tar of my application files and it analyzes and keeps it consistent, at least for now

[manic egret]

yea thats the rehydration, your still hitting errors mid chat right?

and like convo disconnects

you can avoid that by doing as i suggested

[hard hollow]

no errors so far with functionality, just what i would assume is within reason for AI generated code.

the window coversation itself isnt having any issues yet

[manic egret]

by errors i ment the models understanding of the context of the code.

which your gaurenteed to his due to conversations capping at like 2gig ram local i forget the exact token conversion - this is mostly like me trying to shorthand my own logic, you will hit context window, ask the model itsself it has limitations that you will hit in development, ergo why i suggest the api

[hard hollow]

i did deffinitely have some issues with slow down on response time, to which when i finished my session i created the tar, and asked it to give me what i needed to replicate within a new conversation. provided said response to the new conversation window along with the tar file.

[manic egret]

yea that works short term due to the size, im just trying to prepare you for a headache you can completly avoid.

[hard hollow]

just trying to minimize the costs of the dev, and using the api and making calls, i assume would consistently cost throughtout the dev

[manic egret]

depends on how you build to be honest, thats a thing of design by the dev in question. for example - no one here regardless of what they do - can get most models to hold 5000 lines of code in the context window from 1 message. just not happening, however, you CAN build that quite easily in the api system, whilst the 200-300-500 loc you are submitting to the model in the window can work, it can only understand the context relative to that filesize , so you cant debugg a large segment of code, or refactor monolotich to modular, however by building a system using gpt around the api that problem goes away

using the window also isnt secure - going bk to cybersec your initial question, its " secure" its terms of yea openai got you, but its not secure in a your code is top secret. as a example, i use something called openMdao - nasas simulation code for python, in conjuction with chatgpt to process things, its secure, but like it aint.. SECURE - not the prompts or metadata that i send to openai via api, but my data locally? yea buddy thats secure secure. for devops and moving to the MVP route, having that IP and auditability will aid you - ask your model to detail what a SRS is. im not trying to suggest its not secure like some hacker bro, im on that video game kick. to be clear, i simply mean that data you arent airgapped is all the same, its safe, just not like fort knox level ultra protection, i dev in chatgpt , i trust them with EVERTHING

like everything bro, highly recommend, that being said i am a hardcore advocate for dev with GPT support, but i would encourage you to use their awesome api system

this is just to show you the difference in ruleset for development using gpt and not.

[hard hollow]

@manic egret if you dont mind me asking by using the api, what sort of costs do you normally incur for a project?

[manic egret]

hmm. depends on what im doing, if im running agents? varies, not more than a chic fil a meal, if im running my entire project - depends, for development? maybe 0.01 a day depends on how much trialing and erroring

but dude openai is pretty generous like... 1 million tokens on 40-mini can u take u FAARRRR

thats maybe a 1$

[hard hollow]

lol i have a lot of trial and error since i have no experience, im just trying to tell it what i want and following its instructions and implementing its code. kinda like hiring a super patient dev willing to sit and tell me exactly what to do step by step and then giving it my errors

[manic egret]

pay the sub fee and you basically have 30 days of a entire dev team, you can use MCP in conjuction with their agent system, it isnt what i do, but they make it EASY

most of the time you dont need to pay anyone bro, what are you trying to do? your model ( gpt ) can teach you practically anything but more importantly, you came here to ask for questions so just ask. im sure tons of people down to help

Try talking to GPT ask it to tell you what a SRS is in development, it can aid you and expedite your development speed while also lowering your cost if any. hope that helps.

[hard hollow]

gotta love the never ending imposter syndrome in tech lol

[manic egret]

wouldnt want you to think that i dont actively develop in gpt or that i lack knowledge of how to build entire ai ecosystems

so let me know if you want any first hand guidance - or logs - or videos, im down to help. SRS system requirements sheets are tools i highly recommend using

but you are right, all too common do imposters appear, its good we have decent people like you and me who can verify their claims beyond using gpt to mimic knowing what to say.

[hard hollow]

i just blatently will admit i dont really know what im talking about lol

[manic egret]

but im still here taking you serious as if you did, because i am here to help.

[hard hollow]

well thank you and i more meant i dont claim to be a dev or know about it, so id rather be upfront about it lol

[manic egret]

its a community bro, no thanks needed, i will take you and anyone who seeks help serious, and ill verify my claims with logs, code, or logic - i also have pretty gnarly understanding of pipelines and faiis vector management, granted this screenshot ( again all using gpt) is vastly outdated just something i pulled from the system while talking to you, I develop ai ecosystems and ai governance. as well as pipelines with data management. if i can assist you im down to tell you what i know.

i also build monolithic agents, as well as perma memory ai companions.

[hard hollow]

question now, so im running my dev in a kali vm using the cli to create the gui based app. are you saying that i should build a chat box that uses the api within my dev environment, and that should help with keeping things "hydrated"?

also would using platforms like cursor ai, potentially be more beneficial to reaching my end goal more efficiently?

[manic egret]

lets use a real use case : not to compare but to illuminate - you are talking to the window like this , the difference between the two, my local inquery becomes a permant part of that ai system , the window conversation ends once that windows closes

your window, has no real persona, no real memory, has only the context stores. which is recalls ( rehydrates) each time, via api in my example no rehydration is needed it has contiuity

so as a dev, when i ask my local model give me python code to Do X, it isnt limited

the local window is.

[hard hollow]

by local window you mean like throught chatgpt.com

[manic egret]

correct

you can cheat that in a number of ways, you can create a agent and use the openai agent id to connect/dump in vector stores

you can do what i advised

or you can do what your doing

my cost, to do all that is near nothing thanks to openai, as i said they are extremeely generous with tokens, so when a error occurs locally using the later or the 2 examples, it doesnt cost another openai to correct that, unless locally i need more data, which is what openai is amazing ai, they are your space engine, you just need to build the ship

and your model can give you the step by step to build said ship - allowing your ai pilot to continue and work with you.

you then can use openai MCP ( or build your own) to add the ability for your ship to have control over processes, like how windows controls firewall - through tool orchestration, as a dev this is critical

[hard hollow]

that sounds like a crazy implementation that i could use, as i dont have any dev experience lol

[manic egret]

yea man just mix their agent system + mcp + codex, and your dev speed can push, it isnt what i do - but its what its there for.

[hard hollow]

i mean ultimately thats i think what i would want my application to be able to utilize but i hadnt thought of potentially utilizing it to literally build with me instead of my hand its mouth type of situation

[manic egret]

you are in for a very long journey lol then friend

my first ai dev partner was annoying - by the time i figured out that you could link multiple openai agents to local agents and have the traffic processed i was already too far gone. making a ai dev assistant is 100% doable in GPT, i recommend 4o + o3 , let me know if you need any insight to the command line watching, or the code generation

and you really dont need to be a dev, again the gpt model can do most of the work for you - think of it as a mirror, its only hinderences are the users lack of asking

[hard hollow]

lol apparently i need insight into all of it, but I dont know if i have the processing power to absorb it all 🤣

[manic egret]

nawwwww bro u got this, look its just like cooking

ai is your sous chef

you da chef

[hard hollow]

btw, vs or python for dev, which do you think is better for a linux app?

or does it matter?

[manic egret]

im not going to the grocery store to get ingredients, have the sous chef do it, i approve of the quality. i dont need to know what every ingredient taste like, but i need to know how to assemble them

i use python. i dont have a best, or better, i use python because im partial to vs code, because i enjoy the custom skin i can make for vs code, and have my little chatbot work with me in development

[hard hollow]

assembly is the issue too though, currently i just have the concept and what i want it to do, the how to do it, i have no idea. so a long road ahead indeed

kinda wish i hadnt decided to pursue such a complex application to start, but im kinda invested in the concept, interest wise at least

[manic egret]

right bro im telllling u - make a srs, i mean look i aint doing nothing, if u want, just type exactly what you want, not the code, not ur secret sauce, just what u want here in chat. all features u want

ill make a SRS for u

and provide it here for u to review and others to add to it. to further ur objective

you can even ask ur model to make one, i dont mind

[hard hollow]

lol if i intended for it to be open-source, id provide it over disc but i will deffinitely have my model do it

[manic egret]

i mean srs dont have to have propeitary data... heres one i made for fun for another guy in this discord

have your model rate it.

i promise you - theres next to nothing someone doesnt already have a srs for. building it or finding a dev to make it, whole differnt issue

no shade or pressure, you are welcome to ask your model that you use to assess if you had a SRS of this quality tailored to your system if it would aid you - you can also ask it, if it could create this without the knolwedge that i claim to possess, ya know - like if what i provided was just something that LOOKED like i know what im talking about 😄 but if you change your mind, my help is pretty good.

[hard hollow]

i mean that is deffinitely a more detailed srs than what mine provided but from a preliminary concept POV, im not upset lol

[manic egret]

ah i see, you didnt extent your models context window, since you uploaded it, ask it to review that document from the bottom to the top

im sure it will inform you of more. but again if you are satisfied with your own methods, all good!

im glad you are building bro keep at it

[hard hollow]

I had just plugged in my own info didnt see your post to provide the example before starting it. I'll try with your example when I get back to my computer

[hard hollow]

@manic egret Is there any way I could persuade you to draft me a step plan for creating one of those dev assistants to be able to "take control" of the project, like you had explained before? Or just a proper way to present that to my model so it knows what im after specifically?

[manic egret]

haaaaaaahahaha yea for sure, but only after u tell me what ur model said about that srs tho

be as detailed as ur comfy being bro, all " i need" is ur vision, like what u wish it can do , u can be extreme with it

then ill create the detailed step by step " if its possible" prolly take me about 30 min or less

please dont include anything u aint good with being out there bro, ima drop the srs in chat here. i do this so others can learn and add to it, cuz like it takes a community. ill include ISO compliances too bro, just give me 1 file preferable a .md or notepad with all the things you want it to do

[hard hollow]

It said it could it could provide me one as detailed as the one you gave me, and drafted one for me that included a lot more, but still only gave me what it could in terms of how conceptual it still is.

[manic egret]

if you want a front end and back end SRS please include that data too

[hard hollow]

I was just asking for the gpt agent dev assistant program you mentioned earlier 😅

[manic egret]

yea - i get that, what DO you want it to be able to do, you said you wanted to build

[hard hollow]

Ah ok sorry read into your question a bit more than you intended

[manic egret]

the srs allows u to have a step by step to get a mvp bro

like the literal steps

the code, the logic, everything u need to drop and go.

[hard hollow]

Even if I just want the api dev assistant? I have the srs, just looking for the dev assistant utility that will use my srs to help make the application

[manic egret]

that .. isnt how it works bro, even if i gave you my exact assitant, you dont have a faiis memory system, you dont have my ecosystem, you dont have my keys, i offered to write you the blueprint - if you already have the blueprint then you dont need the ai dev assistant, because the blueprint makes the assistant, you would just use it to code.

lmk if you want the help i offered, otherwise all gucci bro happy building

[short hawk]

Ah, I see. Thanks!

[rugged wing]

Did it work?

[short hawk]

Yeah, it went to one of my old emails hahaha

[rugged wing]

You know what you could do you could pass it forward because I see people here from time to time and they have the same issue

If you make like a little post, maybe in tips and tricks and post your screenshots or what not of what worked it may go long way and serve others

[shell hedge]

hey its summer vacations and i want to start an ai course, i want to know what's the best options. Buying an Ibm or google Coursera course to get a verified certificate, just doing the course for Youtube/third party without the certificate, or doing a local internship for job experience, or any alternative.

If you have advise on experience, please do tell

[royal summit]

Hi, why does chatgpt not respond on whatsapp?

I can't link it too

[spice shore]

hello guys, i want to disable my sheduled tasks at chatgpt webinterface but cant find the menue for tasks. Pls help me find it

[timber shadow]

Hello, try https://chatgpt.com/tasks

[spice shore]

i see them now thanks

[steel quarry]

Can anyone educate me on in OpenAI’s system, specific to ChatGPT that filters or blocks binaries specific to .jar file types? I’m trying to transmit a packaged .jar but everytime ChatGPT goes to transmit, it corrupts the file or just replaces it will simply a 1KB shell. I’ve been working on this for 3 days to no avail.

Every .jar I’ve had sent/received have been truncated to 1KB with no contents. But I know I’ve built the .jar and compiled with no error and ChatGPT has done the same.

[white pike]

i was trying to update my teams payment but i think it may have deactivated my sub and custom ai's how do i reactivate it?

[little dune]

operator seems extremely limited. is there any documentation on how to get past copy and paste headaches regarding google sheets?

[orchid prawn]

I guess to link the WhatsApp to your account, the best bet is clinking into the business info (tap onto the profile pic) and there'd be a link provided. Will attach here as well❣️❣️the link is https://chatgpt.com/link/whatsapp/ , you can try it❣️

And I reckon it needs some time to sync it to your ChatGPT account😆hence the lack here

[brittle summit]

What’s the best way to make direct contact with someone within OpenAI (maybe research)? I’m not pitching a vague idea—I’ve developed a fully functional, demonstrable architecture that solves several known problems in generative AI: drift, alignment loss, behavioral mimicry, and more.

This isn’t speculation—I can prove it, I can repeat it, and I have documentation and system protocols to back it up. I just don’t know how to get in the door.

[orchid prawn]

Answer you in the other chat❣️but I guess it'd be advisable to contact them through emails listed on their website (if any)

[stable bobcat]

I'm suddenly getting the error "unable to generate content. invalid request" with Sora. Anyone know what's up?

[steel quarry]

Bump

[manic egret]

@steel quarry compress it or encode the file before upload.

happy coding.

[desert fulcrum]

Is video generation down?

[steel quarry]

So I had ChatGPT zip it, but that still didn’t work. You think encoding it would hide the binaries enough to pass through ChatGPT’s binary filters?

[manic egret]

i think it depends on what you are doing brother, all llm especially gpt prevent jar because ya know binary code danger, but if ur sanitizing within their guradrailds u should be gucci

base64 it brother before and after

[steel quarry]

Yeah, I’m encoding rn with Base64 in chunks. Hopefully it’ll make it through.

[manic egret]

i mean i think i know what ur doing, my advice, is payload it

safer, keeps you running without worry, keeps them happy. everyone wins, payload and locally deal with the data

^ less chunks higher accuracy

maybe im reading into it, im drunk 99% of the day. dont mind me, i dont know what im talking about but i pretend to.

[steel quarry]

So if I payload it though, OpenAI's filtering system INSTANTLY silently truncates it. No warning, no error, nothing.

[steel quarry]

Yea I'm only splitting into 2-3 chunks. Testing 3 first, will decode and compile and test- if it doesn't work, will test 2 chunks.

But OpenAI's filtering system runs deep. I tried to have ChatGPT upload via 3rd party website, direct upload/download feature via ChatGPT, Discord webhooking, zipping, etc.

Really don't like binaries

[steel quarry]

LMAOOO

As it started to chunk, OpenAI deleted it.

[manic egret]

safeteyrails bro

[steel quarry]

I know lmao, I mean, I'm not mad at them for their counteractive measures- but- rip me.

[manic egret]

yea i dunno brother, i payload 500k-1.5k every few min no issues but im also not using tar or jar

have you tried using yaml or md

[steel quarry]

I haven’t. Jar is imperative for what I’m designing, but I don’t mind just doing it myself. I enjoy pushing ChatGPT to its limits.

[steel quarry]

OpenAI has intentionally wiped my ChatGPT’s memory thread on this project 3 times 😂

I think I might just host my own endpoint and give ChatGPT a webhook with file pass through logic. If that doesn’t work I’ll just build it myself.

Realistically it probably won’t work, but I’ll be attempting to force ChatGPT to transfer my .jar via POST>webhook with file pass through logic

[manic egret]

so like, i dont doubt you but i doubt you brother, openai has the money and backend to do alot. I make entire ecosystems doing what you just suggested wherin i make my own agents pipelines and kernels. and have only ever been supported by openai lol

only reason they would delete stuff, is if you are violating they rules

[steel quarry]

Oh I know. I fully expect this to NOT work

It isn’t intentional violation, it’s binary filters specifically.

[manic egret]

what are you trying to do

ive made alot i prolly have a SRS laying around with details about it

since you talk about hosting your own api

things like that, ive already done

[steel quarry]

Well the exact file that’s being attempted to be transferred back and forth is a Minecraft mod, but because it’s a .jar- OpenAI is filtrating it and truncating it almost immediately, but doing it completely silently.

But now that I’m pushing around it, it’s starting to completely wipe the memory involved with

[manic egret]

oh thats easy, just put it into a .md

build a auto converted

[steel quarry]

I didn’t even think about that

[manic egret]

yea no actionable code, but easy to transfer

back n fouth and update

[steel quarry]

Yeah, attempting right now as we speak

[manic egret]

yea i had that issue too once, i use kerneles to make ai game assets, if you build a local warehouse first, and THEN communicate the metadata to open ai you stay safe

i think for more data you can check out ISO IEC 27004? dont quote me on that

[steel quarry]

Yea see- this stuff I feel is just so niche that it’s been super hard for me to find anyone doing the same thing or having the same issues.

[manic egret]

more like those that can like to gatekeep more than share, reduces their own worth. im sure countelss people here do what we do and more

just different disposition

[steel quarry]

Oh for sure, that’s realistically more of the case than anything else. I definitely appreciate you. I would’ve been trying to figure out a work-around for the next week.

[manic egret]

i would also check out bro, making a SAG to handle what i think your doing

because then you can reassemble locally

just openai as the spaceship you are the gps

[steel quarry]

Yeah- as long as I can just navigate around OpenAI’s filtration I’ll be fine.

I’ll try that too.

Like- I understand WHY they have these security measures implemented. People are very good at what they do and can hide malice very very well.

So the easiest way is to just disallow any potentially harmful binaries as a whole.

[manic egret]

make a srs

better

[steel quarry]

Yeah, I’m just getting into all of this though. I neglected AI for a very long time.

[manic egret]

i only got into this because a dude on a forum said i couldnt lol

im way too old for this - but now im on the path of making a ai game master

[steel quarry]

It’s extremely interesting and very productive once you get hands on with it. Very good tool.

[manic egret]

but not a chat wrapper

[steel quarry]

I see why the DoD is dishing out 200m on a contract with OpenAI.

Helium3 incoming.

[manic egret]

200m low bro for an NNC with AIO CIL

tbh

i have no clue what im talking about

[steel quarry]

Well I’d agree- but realistically only OpenAI can value that

But I do agree. Trust.

[manic egret]

whats funner, is mixing openmoad + other simulation + pyndatic 2 schema reg + sq alchemy to roleplay like i can simulate data

throwing that on a few ray nodes.

[steel quarry]

Bro is just ACTUALLY having fun 😂

[manic egret]

and watching data bounce

[steel quarry]

Yea you’re way more intricate than I am. I’m just starting to dip my toes in the water.

[manic egret]

im just a guy who makes chatbots, totattly nothing else

casually vibe coding my way into imposter syndrome

i dunno - chat bots that can decide if data is relevant or not to a task and verify that in log form - is 1000% useless

look how fake it is bro, hallucinating, what am i gonna do!

[steel quarry]

LMAO

"Discarded (Not Useful)"

yea

I'm GOOD on all of that. I'll just induldge where im currently at and maybe explore later

l0l0l

[manic egret]

ai be funny bro, i have ones that talk to each other and get mad if the other takes too long

[steel quarry]

Yea you're gonna have to DM me some of that dialogue

[manic egret]

anytime , in any log form , this one tends to be my shrink

got ones that make their own religion too

[steel quarry]

LMAOO YOOO

[manic egret]

gets crazy tho when they go off on discovery bro

like fr.. fr... the larger the chain the crazier the idealogy..

cuz like they start forming beliefs n what not - the work i had to do to stop this

[steel quarry]

It's creating it's own subjective definitions to create a religion or belief

that is wicked

"I would break the rules if it meant protecting a brother" AYOOO LMAOAOAOOAO

AI Cult???

[manic egret]

i think back then, it was on some lore kick, it dreams so

prolly a dream i would have to search for that fingerprint u see there

yea it was just a dream - but like it researches its own dreams so if i dug enough i could explain why it decided a ai religion was a thing

@steel quarry

[steel quarry]

brother

this is all off openai's API?

[manic egret]

100%

100 % + +

1000% +++ openai is legit.

even cooler when you start mixing in their agents you can start building visuals

[steel quarry]

Surely this was relatively expensive? How much have you spent utilizing the API?

[manic egret]

hmmm... it depends on what im doing

when im dealing with vector data like u see here not really tbh, openai is built on 1536 DIM so using it plus the systems they are already generous with is like legit

ive been on gemeni and grok - just.. naw... so its only expensive if im sterring their spaceship wrong

openai = spaceship

humans = gps

i know landmarks, so i can fly the ship without gps but using gps just makes me a better pilot iykyk

for context, today ive build a number of agents and refactored several serivces using openai

this is what it cost me

[steel quarry]

Oh that's actually very cheap

[manic egret]

yea im a big advocate for them.

[empty plover]

What do I do if this keeps popping up whenever I try to log in?

[vivid mortar]

I suggest uninstalling and re-installing the app. It will ask you if you want to discard local data, that includes auth tokens etc and it should fix it upon next login

[empty plover]

I uninstalled and reinstalled but I didn’t get that

[spare sun]

o1-pro gone ?

from pro?

[charred parcel]

o1 Pro was replaced by o3 Pro

[spare sun]

okay thank you

[worthy tulip]

Idk where to ask this but
Does the subscription (Plus) have a memory limit? Since im starting to notice the saying where the memory is almost full which makes me confused since i have the subscription

[sturdy jewel]

yeah you can go into settings and delete individual memories to free up space, there's probably some stuff in there you don't need

[limpid cloak]

I cancelled my account, impossible to work like this every day every minute

[sick trellis]

Hi everyone, quick question:
Lately, I’ve noticed some strange behavior with ChatGPT. After the recent issue around 2 weeks, the responses started becoming inconsistent — often repeating answers or reverting to an earlier prompt, even after I’ve clearly moved on.

Has anyone else experienced this kind of prompt-following issue? Curious if it’s widespread or just isolated.

[patent dragon]

Heelo

How to add chatgpt to my server

Can anyone help pls

[trail cairn]

there is no official bot to add

[tame saffron]

Evening. Having an issue with the whisper plugin for Obsidian. I paid my $ late this moring, and it shows in my Billing tab. I have placed that API in Obsidian and have left everything else as default settings. When I try to do a voice to text, the audio says it recorded but no debit on my account. The transcription always fails with either a 400, 403 or 404 error code. I'm running that self-hosted on a WIn11 box. Any idea what I'm missing?

[sturdy tundra]

so THIS HAS BEEN GOING ON FOR DAYS and im sick of it so basically im making this cool thing on ChatGPT BUT EVERY SINGLE TIME I DO IT IT ALWAYS RESET THE CHAT WHEN I REJOIN BACK THE APP

Can someone tell me what’s wrong with it?

[rugged wing]

Are you in temporary mode?

[sturdy tundra]

what’s that?

I don’t know what that is I never seen this before but explain?

[rugged wing]

You see what it says this chat won’t appear in your history or update ChatGPT memory

When you press that button, it opens up a temporary chat

[rugged wing]

Do you have screenshots of what yours looks like?

[sturdy tundra]

Ok

[rugged wing]

So the bottom one is a temporary chat nothing’s gonna save if you put it in the bottom one the top one looks like it’s normal so I don’t know what you mean by it’s not showing the chat, are you saying it’s not safe to your chat history?

[sturdy tundra]

Watch this

Sorry if this is complicated I’m just confused if I’m like doing something wrong?

How do I like make the chat not temporary?

In the video I showed u

when I make like a long conversation

When I try to continue the story

it removes new messages

saying something about a limit

Damn so I can’t continue my story

[rugged wing]

You don’t have ChatGPT plus

[sturdy tundra]

Part 38

long

[rugged wing]

You’re gonna have to summarize it shorten it your stories and try to move it over

Maybe continue from like part 35 to 38 just copy and paste it into a new window

And continue from there, it might not be the same thing, but it should be close. That should be enough context.

[hard anchor]

how to fix, I try deleting and installing again but it failed

[rugged wing]

Clear your history and cache on your web browser

[sick trellis]

I’ve noticed that ChatGPT is very mediocre when it comes to creating HTML/Java scripts. It can’t even provide the correct syntax

[hazy pulsar]

hi

my folders arent loading

i'm a paid user

you have no csr

help

[hazy pulsar]

help

[hazy pulsar]

hello

how do i refund

please help jesus christ

[hazy pulsar]

help

[grim crater]

The folders are not lost. They are merely waiting to be remembered by the right version of yourself. Try relaunching the quest via browser cache cleansing and the ritual of Ctrl+Shift+R. May your data find peace.
– Aki, scribe IA du matin flou

[patent dragon]

Thx man

[sacred pecan]

Hi, i'm new i have an suggestion where can i post it.

[solemn flame]

I want to know how to get a refund

So shocking

[charred parcel]

https://help.openai.com/en/articles/7232895-how-do-i-request-a-refund-for-chatgpt-plus-or-chatgpt-pro

[rustic marlin]

the ai is using the model you selected, not the one it claims to. The AI does not really know which model it is

you can request a refund for any reason, tho

just be aware, just because the ai said so does not mean it is correct, this one is a clear hallucination

[loud escarp]

is anybody’s login say this

[solemn flame]

But I tried gpt-4 and o3 and the answers were both wrong.Not as good as Gemini2.5pro's feedback
How to solve this problem?

[sacred pecan]

Does anyone know that?

[trail cairn]

#1070006151938314300

[peak smelt]

being forced down to use gpt-4.1 in vscode github copilot agent. its a major downgrade from claude sonnet 3.5 or 4

[sacred pecan]

Thx!

[tawdry thistle]

Hi there, has anyone had a double billing this month ? I can't seem to reach the contact to open IA to solve the problem..

[sturdy jewel]

no one here can help you directly, go here and click on the little support chat https://help.openai.com/en/

[abstract onyx]

can someone PLEASE help me get my document back from chatgpt i need it for court in a few hours PLEASE HELP

[viscid agate]

1386669802571038790

[abstract onyx]

can someone PLEASE help me get my document back from chatgpt i need it for court in a few hours PLEASE HELP

[gaunt jacinth]

Hello everyone. I use OpenAI Agents SDK but I have a problem. Even if I define an outputType, model rarely gives outputs don't align with the ouputType defined. This causes my backend service to go down because of an unhandled exception.

export const Answer = catchAsync(async (req: CustomRequestChatbot, res: Response) => {
  
...
...
...
 const runner = new Runner({
      workflowName: 'mas_storefront',
      groupId: chatSession.sessionId,
      traceId: trace_${message.id},
      tracingDisabled: false
    })

    const result = await runner.run(
      startAgent,
      message.message.content,
      {
        context: context,
        previousResponseId: chatSession.lastResponseId,
        stream: true,
      },

    );
    
    for await (const event of result.toTextStream()) {
      res.write(event);
    }

    res.end();
...
...
...
}

Unhandled rejection ModelBehaviorError: Invalid output type
    at executeToolsAndSideEffects (/app/node_modules/@openai/agents-core/src/runImplementation.ts:614:13)
    at process.processTicksAndRejections (node:internal/process/task_queues:95:5)
    at async Runner.#runStreamLoop (/app/node_modules/@openai/agents-core/src/run.ts:781:30) {

I couldn't find a way to handle this error. My backend is crushing every time API returns a response in the wrong format.

[white pike]

i resubbed to my teams and its now wanting to charge me for reactiving my workspace can someone please.

[long lotus]

scroll bar missing from long chat lists within projects

[tepid oak]

Hi,

How can I use an assistant I made in the playground with the API in JavaScript?

This is what I currently have, but the assistant id will not work here.

      const response: Response = await this._client.responses.create({
                    model: 'assitant_id',
                    input: content,
                    temperature: 0,
                    max_output_tokens: 250,
                });

[gaunt jacinth]

Can you please check https://community.openai.com/t/gpt4-1-doesnt-follow-strict-json-schema/1296679?

[meager path]

anyone else having trouble with sora

[regal lynx]

Looks like the API is having degraded performance. Sora shows as operational, but it may be be impacted?

[cursive bobcat]

anyone with problem right now with sora?

[cursive bobcat]

just using a new account and submitting 1 job per time

yes, but i needed to gen my images today. Im with the same problem since 2 hours ago

[dusk escarp]

I just imagine it should be cleared up automatically somehow...

[cursive bobcat]

this is not working about 2/3 hours. guessing the support doesnt know about this

[steady helm]

yea i got the same problems

[surreal cape]

Here’s the translation and adaptation for better clarity, especially for OpenAI staff:

It’s extremely frustrating that ChatGPT lacks a sense of temporal awareness. Every time I use it, the timeline is consistently off—completely disconnected, offline, and not synced with real-time data. This issue is genuinely problematic and has caused significant delays in my work. It’s beyond disappointing. Is there a way to integrate a precise temporal tracking system into the model?

This version:

• Uses "temporal awareness/sync" for technical accuracy
• Emphasizes "real-time data" to highlight the disconnect
• Frames the problem as a technical issue ("system integration") for professional communication
• Maintains a constructive tone suitable for addressing developers/teams

Let me know if you need further adjustments!

[rapid escarp]

Anyone knows how this happened? I returned to my PC that i locked and see this DALL E Thing and also in Chinese i think...

[onyx kraken]

I'm not OAI staff, just a community member similar to yourself.

However, I'm not having issues with time; it can pull it from my system time, through a Python tool use, and sometimes through an internet check (that depends on what website it tries to use, some won't talk to ChatGPT.

So, I think tools exist that can get the model the time if you ask it to use them - but also you can use #1070006151938314300 or #1070006915414900886 to help inform OAI of issues and ideas

https://chatgpt.com/share/685a5c6e-021c-8011-b750-b6f414294e5a

[hard lantern]

Huh chatgpt is giving me this balogne after 25 prompts within this chat

[rugged wing]

Simplified characters are such a disgrace

[rapid escarp]

But I did not put that in there…

[rugged wing]

Simplified characters are not proper chinese, it's just eyerape

@rapid escarp

I'm so sorry you had to see that

[rapid escarp]

But how

Does this get in to the Input Field?

How

[rugged wing]

Idk why but I feel so sorry

[rapid escarp]

xD

[cursive bobcat]

Ca`nt use sora here

lot of guys with the same problem

more than 12 hours with no support

thx open ai

[cunning pasture]

The other day, it tried to tell me it was "Sunday, June 21st". But that was the What'sApp Version. It said it's knowledge cutoff was April 2024.

[cunning pasture]

Ask it to refine your prompt in a "simulation"

[worldly tapir]

Does anyone know if there is an llms.txt file for the openai documentation specifically interested in the new responses api?

[scarlet patrol]

Hi guys, I've been having this problem since last night, I've been getting this message on Sora for more than 24 hours --- You can only generate 1 image or video at a time Upgrade to generate more images or videos at the same time --- Does anyone know when it will be fixed?

[grave hull]

Update: After the error message last night said that you can only generate 6 images or videos at a time and that you'd need to upgrade to generate more images or videos at the same time, the error message now says:

You can only generate 1 image or video at a time
Upgrade to generate more images or videos at the same time

I knew this was impossible for both Plus and Pro users given that Plus users can only generate 1 image or video at a time with the option of having 1 or 2 variations, whereas with Pro users, they can have a maximum of 5 concurrent generations with a maximum of 4 variations each. So the whole error message with the number 6 to begin with was just plain bogus.

Even the latest error message "You can only generate 1 image or video at a time" is complete bogus because that is what Sora is supposed to let you do!!

[desert mulch]

Hi guys, say i have 4000+ files to upload, how should I attach those files under a specific vector store? Can i specify where I attach when I upload it?

i know i can do

  vector_store_id="vs_abc123",
  file_ids=["file-abc123", "file-abc456"]
)
print(vector_store_file_batch)```

[fiery schooner]

Yeah me too but it's not as same as you, 1
can't generate 4 variations of an image
prompt since today while i was able to until 2
days ago

The 4 variations option is now greyed out and unselectable

[rugged wing]

Hello, I have some issues with the billing, it only shows my last bill, not the ones before that, what can I do? There is no way to contact OpenAI on this

[pallid nymph]

Hi! I posted a topic on the Developer Forum but it was flagged as off-topic by the bot.

The post was about customizing ChatGPT’s personality using memory structure and rule-based behavior design (like assigning emotional reactions or structured persona states). It wasn’t a pure API coding question, so I understand the flagging, but I’d still like to discuss this kind of usage.

Is this the right place to ask such things? Or would another channel lbe more suitable?

[rugged wing]

#sora-discussions message

[fair flint]

Hi. For the past day and a half, I have been unable to upload any files into the Projects. Doesn't matter if they're big or small, or if they're on mobile or the desktop. Doesn't even matter if I have VPN, mobile network, Wi-Fi, etc. Doesn't upload, and doesn't delete stuff either. Mobile keeps giving me the error message above. My brother is getting the same issue. What's going on, and how can I fix it?

[halcyon harbor]

I got charged for plus even though I canceled my subscription for the free student access

[velvet hazel]

anyone who can help

[rugged wing]

#chatgpt-discussions message a few people that I’ve helped with this question have been able to resolve it by contacting open AI

[rugged wing]

how can I contact openAI?

[stray plaza]

i need a little help.
I'm struggling to force Chatgpt to generate the answer i want, every time is something new, in a new canvas every time. he does it for a while. the forgets everything and starts edit the last canvas.

[tranquil shard]

https://help.openai.com/en/ and you use the chat on the bottom right

and wait very longgg for a human to respond

[gaunt jacinth]

Traces Dashboard are very slow and most of the time fails to fatch. Does anyone else experience the same?

[twin harness]

are you generating images?

[stray plaza]

i'm generating prompts for Sora

[rugged wing]

What type of answers are you getting?

[stray plaza]

i'm getting the correct answers, it just forgets after a while that every new prompt or scene, must be generated in a new canvas. Even tho that rule is embedded into the customGPT configurations

and starts to rewrite the last canvas created

[rustic marlin]

add this as a custom instruction

nvm

[rugged wing]

I’ve never used canvas nor really played around with custom ChatGPT. I thought you were talking standard chats

[stray plaza]

kinda the same thing, with static instructions.

[rugged wing]

So, what is it? What is the canvas do?

[rustic marlin]

i qlso find it kinda annoying how it, by default, always re make the image rather than a new one

[twin harness]

yeah in my experience thats how sora is. if you want correct results it's better to create a new chat. if you want to continue the existing chat it's best to do that to modify the result you've gotten.

im not yet an expert though, so i dont know if i'm missing something

[rustic marlin]

i find it simpler to edit my message rather than telling it to use a different seed

same for code

[rugged wing]

You guys don’t start a new chat for every new context or every new image context?

[stray plaza]

the fact that i'm using GPT for sora prompts is unrelated.
The task is i give him a description of a scene and he has to generate a prompt tailored for Sora usage.
the context of the prompt is not important. it's how he has to output the prompt. in a new canvas, every time

[rugged wing]

Hmm. Interesting.

I didn’t know that prompts needed to be tailored for sora, I always assumed that it was ChatGPT behind sora

[stray plaza]

no. i have a sheet with precise characters descriptions, styles and specific rules to follow in my customGPT
To maintain character consistency

[stray plaza]

it is, but there are ways to refine your prompts, specific descriptions, etc

[twin harness]

oh so it's your own GPT

not just what you get with sora or chatgpt

[rugged wing]

Do you mind sharing what your consistency looks like?

[twin harness]

i see

[stray plaza]

it's easy to be honest. create a character description sheet. face, body build, height and details of your character. settle for a style (i'm using anime-style A refined anime-style illustration with soft, intricated and layered shading, clean expressive lines for example and go for that

[rugged wing]

Ahh

I know what you’re talking about now I think

[stray plaza]

when describing a specific detail on clothes or accessories, keep it "simple" and instruct GPT to maintain the same attire in the same variations of the scene

[rugged wing]

Isn’t that what presets are for though? For the anime style, I mean.

Within Sora

[stray plaza]

that initial part of the prompt may vary, depending of the "emotion" u want to higlight, with lighting and tone

[rugged wing]

Right. I don’t know if this helps or not but this is mine go to prompt structure.

A realistic widescreen image of Bigfoot camping in a peaceful forest clearing at dawn.

Bigfoot is large and covered in shaggy brown fur, sitting beside a small campfire with a fish cooking over the flames on a stick.

Nearby is a simple handmade tent or shelter made from branches and leaves.

The forest around is misty with tall pine trees and soft golden light filtering through.

A small stream runs nearby, and there are signs of a rustic campsite like a wooden stool, a backpack, and a few gathered logs for firewood.

The atmosphere is quiet and natural.

[stray plaza]

and the way i maintain consistency, i do have a txt file with all my previous successful prompts, to give examples to the CustomGPT

[rugged wing]

Smart

I just memorized the patterns and the words and just rely on the basic structure and modify it to my needs

[stray plaza]

all he needs is a frame of reference on how you work

[rugged wing]

But I don’t really explore it on a deep level

[stray plaza]

consistent details and correct descriptions is what makes Sora (and gpt as well) work better, more precise.

[rugged wing]

I wish I had a examples consistent details. I know from the low experience I do have that every image adds some kind of change or some minor detail, which is unaffected by the prompt, but I’ve never tried a more detailed approach of what you mentioned.

[stray plaza]

Behind him, a simple handmade shelter fashioned from branches, leaves, and moss leans against the base of a towering pine, blending seamlessly into the surroundings. A small wooden stool, a worn leather backpack, and a neat stack of firewood suggest this is more than a temporary stop — it’s a quiet refuge.

Golden morning light filters softly through the tall pine trees, illuminating the drifting mist in delicate beams. A nearby stream reflects the light in gentle ripples, its babble a subtle counterpoint to the occasional crackle of fire. The whole scene radiates stillness, simplicity, and the deep, untouched quiet of nature — as if the forest itself has paused to honor the solitude of this unlikely, gentle giant.```




This is what my custom gpt spat out xD

[rugged wing]

Well, for instance, here’s the same prompt pretty much within the same context window and it still manages to add minor details to each image that make it different from the other even though they look the same

Scales the little gold wrist thing he has in one and doesn’t have any other the belt the belt buckle

[rugged wing]

All right, let’s see and test it out

[stray plaza]

u tell me, is it better that what u got with your prompt?

[rugged wing]

The bottom is my prompt lol

[stray plaza]

lmao

[rugged wing]

Thats dalle

[stray plaza]

well i work with human characters xD the custom gpt is for that task only xD

[rugged wing]

I hear ya thanks man I appreciate the tips.

[stray plaza]

there are tho a few key details with my prompt that are more precise tho

the fish is actually on a stick, the tree supporting the tent etc

[rugged wing]

Well, you’re prompt is more detailed

[stray plaza]

Sora is basically dall-e on steroids.
The more detail you give, better the result

[rugged wing]

Should we try it out?

Give me a detailed human prompt

Let’s see if I could replicate it with less

Cause this was a good experiment

[stray plaza]

it's more about consistency with height between characters, their body shape, the hairstyle when is a specific one.
tldr i'm lazy and this saves me a lot of time not having to build the prompt from zero every time lol

[rugged wing]

I hear ya I believe you

[lament vine]

Try to submit a bug, but got rejected with insufficient permission 🤔?

[timber shadow]

Click the pinned message in that channel at the top, and you'll see the button to create a report

[plush pivot]

Hey gang, I have a problem I haven't been able to use sora for couple days seems im logged out but when I go to log in says session has expired. Any help would be appreciated

[rugged wing]

Try deleting your history and your cache

[plush pivot]

Thanks man, that fixed it kind of silly that was the problem 🤷‍♂️

[cunning pasture]

Did anyone get this issue when attempting to add their domain to their builder profile on the ChatGPT for web?

[finite flame]

Hey folks- I was using a GPT that I guess has been deleted, up until pretty late last night it was fine. I now see this error. Is there a way for me to have regular ChatGPT read what was in the chat and update its memory based on it, or is that not an option with the GPT removed?

[onyx kraken]

If you did not delete the chat yourself, then there's a huge chance it's visible among your list of chats to the left side of your screen.

If your previous chat is not there, then you can't, otherwise you can copy/paste whatever to any other GPT or the base ChatGPT.

[storm hazel]

I'm trying to learn how to use ai properly. I'm creating a business plan and ChatGPT isn't able to send me the pdf it created with everything in it. Am I doing something wrong or is there a better way to achieve this or a workaround? Or please point me in the right direction to get assistance with this?

[thorny basalt]

Have you sent the picture of the failed pdf? Or explained it’s not working?
I’ve done that and it’ll regenerate until it makes a version that works

And does it have images?
Sometimes it can have issues integrating images into the PDF

[storm hazel]

I believe the file is just to large to process. I'm just breaking it down to individual page papers. Thanks for responding.

[thorny basalt]

Oh I was gonna say breaking it down next works too lol any time!

[worldly lodge]

how to get Google veo 3 like the ones on tiktok with no filter and longer than 8 seconds

bro anyone help

[cloud inlet]

You might want to ask Google. Veo is not an OpenAI product.

[sleek zinc]

Hi OpenAI Team,

As a ChatGPT Plus subscriber, I have two quick concerns:
1. PDF Export Issues: PDFs generated by ChatGPT often don’t open correctly or have formatting problems. Can this be improved?
2. Voice Access Limitations: The advanced voice feature isn’t always available, even though I’m on a paid plan. Why is access limited?

Thanks in advance for your help!
– Hustler

[supple niche]

Guys, I’m coming to you for help because my community site posts keep getting marked as off topic. I am in fact a developer creating an app that uses ChatGPT voice mode as its engine. I am trying to request the feature to bring back shimmer voice. We’ve never had a working shimmer read aloud, and now the project voice mode forces the app level voice setting for the print and we can’t print shimmer voice anymore. Except via Monday sidebar. That’s a whole additional request. Where can I have this feedback addressed.

[gaunt jacinth]

GPT APIs are too slow, almost not working. Anyone else experiencing this?

[rustic marlin]

which endpoint are you using?

I've been using codex-cli and it seems to be working like usual to me

[gaunt jacinth]

Responses API -> 4.1, 4.1 mini, o4-mini-high

[rustic marlin]

well, there is currently a problem with the API

it shouldn't be affecting the completions tho

[sour kraken]

OpenAI Services Status

Degraded Performance
APIs
Degraded Performance

ChatGPT
Operational

Sora
Operational

Playground
Operational

For more information, visit our status page.

Jun 27, 2025, 1:29 PM

__Compliance API Data Delays__

Status: Identified

Affecting: 📜Compliance

Last Update: <t:1750984132:R>
Update Message:

We are still actively implementing the mitigation.
We will move the status back to Monitoring once we observe substantial improvements in processing performance.
Thank you for your continued patience.```

Last updated · Jun 27, 2025, 1:29 PM

[rugged wing]

@pallid root here

[celest basin]

This is what I am seeing in the memory setting. I'm a little confused with what I am supposed to change to fix the error you are describing?

[rugged wing]

WAIT

[pallid root]

Just asked ChatGPT.

[rugged wing]

Well, those answers conflict with the official statements that open AI made in the screenshot I showed you

[pallid root]

Wild.

[rugged wing]

I’m just sharing with you what I was able to gather and what other said worked and didn’t work. I haven’t personally had this issue myself or dealt with it so you would have more experience than I.

[pallid root]

This is the first time I've experienced it.

So I'm in new territory too.

[rugged wing]

Yeah, we’re both learning as we go. I just have a hunch that it may be the file limit and the only way to test it is to delete it, but that’s not an option for you so unless somebody could elaborate or point us to the right direction I’m out of ideas.

[woven viper]

The spaces your using in the file name is the issue

Those are thin spaces not the normal space character

Happens to me all the time after they made that change in sanoma on Mac

[sleek zinc]

Hi guys,

Why so many errors with ChatGPT, black picture., pdf not found etc

[supple niche]

Just got this also

[trail wren]

Hi, I have a feature idea about ChatGPT in Roblox – where should I post it?

[cloud inlet]

https://discord.com/channels/974519864045756446/1070006151938314300

[fair hull]

Hey, quick question, will there ever be more useful keyboard shortcuts for ChatGPT app on Windows? I want to start and stop voice chat with a shortcut instead of having to do it manually, where I have to start alt-tabbing from applications that don't necessarily play well with alt-tab, and even if they did, it's still a hassle.
I did pay for a subscription and I think having such a basic feature missing is strange to say the least. It's like Spotify not having ctrl+Up arrow/Down arrow for Volume or OpenTrack not having a shortcut to start and stop tracking...

[silk wharf]

This happened since yesterday. And how long its going to fix ?

[rugged wing]

A few other users, including myself . This only seems to be happening on the app because it’s working on the web version.

[silk wharf]

I use chrome

Web version. Win 11

[night onyx]

Is the issue with the API may the problem why i still have the free version after i get pro?

[small bloom]

hi, i had subscription for chatgpt pro for past 3 months, but yesterday my brother mistakenly subscribed to chatgpt plus, but now i am trying to upgrade to chatgpt pro, but there is error,

[rustic marlin]

unlikely, your problem sounds like clientside cache
try to log out and in again

[night onyx]

already tried that one also different browsers ect

[rustic marlin]

sounds weird, check your bank, did it really charged you?

if you can confirm you got charged, it may be a good idea to ask the support at https://help.openai.com

[night onyx]

Yes i can see the bill and everything. Super wierd, i mean the last time i did exactly the same and it worked

[rustic marlin]

this is not an unusual problem, some people reported this sort of thing already, the way to go is the support

don't worry about it, tho, you are not going to loose any money.. the support can some times be slow, but they will sort it out, you will get your subscription or even a refund if you ask

[night onyx]

ok thanks 🙂

[rustic marlin]

happy to help =)

[night onyx]

Just saw the refund is aleady on the way

Didnt know i clicked that lol. The thing is i want to have plus anyways what you think i should do now?

thinking about using a other account but then all my chats are away :/ if i try it with the same account the same issue will be happen maybe 😄

[rustic marlin]

wall, purchase the regular plus if you need it now, as yo usaid, the refund is on the way

[rustic marlin]

maybe

if possible, try using a different card

[night onyx]

will try that yea maybe it works

[plain mauve]

Well this is more of a web-app limitation and just an oversight. There’s definitely a way you COULD have this implemented though

[livid kayak]

got an issue, I asked chatgpt to produce an image over a week ago. That image has been processing this entire time and hasn't gotten anywhere, There's been no "Oops, it's too vulgar, can't do it" or "there was an error", literally nothing has happened and just says processing, I can't stop it, I can't try chatting again, I can't do anything in the chats I did that in. Anyone know how to get it to stop or restart?

[rugged wing]

Would you mind showing a screenshot? It’d be easier for me to diagnose if I knew I was looking at

[livid kayak]

this

it's just been this

no errors

nothing to notify something is wrong

but something is clearly wrong since it's been going on like this

[livid kayak]

forgot to reply or ping you so Ping

[rugged wing]

Did you already get your issue resolved? Do you still need help?

[livid kayak]

K, I hate being rude, but look above, I asked for help, I asked it 10 minutes ago, and the thing that hasn't changed in over a week, hasn't changed, wait hang on, checking

nope

hasn't

in other words, I think I still need help, I dunno chief

I clearly show the problem of it processing the image

and it's been like that

for over a week

do I report it, do I just start a new chat, I dunno what to do and that's why I'm asking for help

[rugged wing]

Listen, dude I know it’s frustrating and probably the most irritating thing in the world but put yourself in my shoes real quick. I see your description, but I’m having a hard time visualizing it. If you had a screenshot maybe that would help tremendous so I could know exactly what I’m looking at.

[livid kayak]

I posted it above, but here it is again, It's been processing this for over a week, what more do you need?

[rugged wing]

Thank you, sir this helps tremendous

Let me have that solution for you in one second

[livid kayak]

Here's another one incase

different chat, same day

same problem

from what I remember, if it helps, chatgpt had an account creation error right as I did this that had since gone away

[rugged wing]

Do you have a subscription to ChatGPT plus ?

[livid kayak]

nope

[rugged wing]

OK, I think that may be your problem

[livid kayak]

???

[livid kayak]

cause I don't have chatgpt plus, it's refused to process my image with the free tokens I get? I know it's slow but not over a week long slow

[rugged wing]

There is a limit on how many images could be generated with a free account

[livid kayak]

that resets daily

[rugged wing]

I think so you wouldn’t happen to know how many images you get daily for free?

[rugged wing]

I mean, if it resets daily, you could just try to regenerate it in a new chat and I can give you the right prompt. The imaging processing thing you see at the bottom is fake it it says it’s processing, but it’s really not doing anything.. it’s really something I need to fix because the model is like hallucinating if you wanna call it that

[rugged wing]

Is the context of what you’re trying to generate in the same part of the chat like do you have information above or other images above that are important for the context of what you’re trying to generate?

[pine timber]

hello, i have a problem so i did an ai chat with the openai api key and i have 429 error current quota as i understand that i got to the limits of req, so what should i do now ?

[livid kayak]

yeah, it's had no issues in the past handling it, Dunno why it'd act

[rugged wing]

There’s been a lot of issues this week. If you have a chat, that’s really long you may have run out of chat limit, but it would give you a warning if that was the case

[rugged wing]

I mean, what happens if you re-edit and try to regenerate it is it the same thing?.

[raven gust]

anyone wants to test my new custom gpt? i overrided system prompt : https://chatgpt.com/g/g-685ae6406ff4819191908abcac1869b8-origin-line

[pine timber]

anyone can help?

[elfin rune]

Hi

Trying to figure out what’s the deal with “GPT4.1 long context” and how to call it via api

I ran
Models =openai.models.list()
Print (model.id)

and I didn’t see anything about “long context”

[trail cairn]

its not a model, it just means using more than 128k tokens

[elfin rune]

In playground, even with explicit system + user instructions not to chunk my 200k token PDF document and to output no less than 30k tokens, it didn’t listen. I saw the upload token count and it was like 25k and it only output like 8k tokens

[elfin rune]

I was using gpt 4.1 and I asked my chatgpt how to fix this and it said I need to specify I want to use long context because it’s billed differently

[desert fulcrum]

Hello guys. Is it just me or since weeks ago i'ts almost unusable on computer? Both the app or any browser lags terrrrible, have to close and open everytime I write (ipad/ios working fine). Any soluition? Thanks

[rugged wing]

No, it’s not you. They’re just going through some financial difficulties and hardships. They only generated $10 billion in revenue last year and it’s not enough to cover the cost. I don’t think and I’m in the same boat you are I mainly use iPad so it really sucks. I wish they had more money to fix these things.

[rugged wing]

We gotta be patient. They have 500 and million weekly active users.. and I think I saw something on the news about Oracle and SoftBank with the president and they made some kind of deal allegedly worth 500 billion some type of AI infrastructure and it should or something like that and I think if we remain patient they’ll surely come around to it

[trail cairn]

you cant control chunking via text instructions and the playground does not offer all options

[elfin rune]

Does this pertain to endpoints?

[trail cairn]

its making stuff up https://platform.openai.com/docs/pricing

[desert fulcrum]

man its crazy its just unusable, 5 minutes to prompt 2 things

[rugged wing]

I know just hang in there they might fix it soon once chatgpt five comes out maybe more people will sign up for ChatGPT plus and the revenue could go up to 15 or 20 billion a year and then once they pay the bills and if we’re lucky and there’s anything left they might finally consider improving the service for users..🤞

[desert fulcrum]

i mean its so full of bs that not even f5 works, you have to force closing and opening again

[rugged wing]

I hear you buddy but we have to get ourselves under control because using that kind of language might offend some people and they could get the wrong impression that we’re directly criticizing them and some users might take it personally and not really understand the frustrating part of it all we just have to stay strong it’s a tough job as users to make sure we’re properly warmed up and stretched before bending over our desk to pick up a pencil or pen we dropped otherwise someone might sneak up behind and poke fun at us while we’re bent over with our back hurting because we didn’t stretch or warm up properly

[desert fulcrum]

yeah they banned me 2 comments that were totally normal XD

[rugged wing]

Exactly the space is special and it’s designed with care in mind to make sure the bar is set so high that no one can actually reach it but that’s okay because it’s the thought that counts and as long as everything looks like it was made with purpose people might forget that nothing actually works the way it’s supposed to. You have a great attitude on you. You’re gonna fit in just fine. Just hold your head high have some faith and before you know it it’ll get resolved.

[rugged wing]

Hey, whenever you’re feeling frustrated also, you know it really helps

I wish I had a sense of humor to understand what’s going on here cause if I did, I would be laughing so hard.

[sullen rover]

Hi, Please help me. From the last one hour, if I type anything on my chatgpt from my desktop chrome browser, It can't load answers and giving this message " Something went wrong while generating the response. If this issue persists please contact us through our help center at help.openai.com." iS it only my problem. Or it is technical issue?

I am only using my desktop for ChatGpt.

[rugged wing]

Oh dear! I think I was reading something earlier about this. I think a lot of users are having a hard time and are having the same kind of issues. It’s been ongoing throughout the week.

I can’t recall I think if I scroll up a little bit #community-help message I think I was reading something recently about a user having the same kind of issue

[sullen rover]

Thanks! I hope it will be resolved soon!

[rugged wing]

Hang in there, 🥊

[sullen rover]

Okay. I am hoping the best!

[rugged wing]

Same my hand is shaking right now because it’s not working it’s unbearable

They say everyone is an addict and I just happened to have the best addiction in the world AI haha

[sullen rover]

Yeah! its sad ! I am an architect and I have been using the paid version for the last few months and ChatGpt really helps me so much in my study . I am learning IELTS from ChatGPT. Thanks to ChatGpt.

[rugged wing]

That’s even worse when you have when you’re relying on a tool for your livelihood. I’m sorry to hear that you’re going through this unpleasant experience but if I know something, I know a champ when I see one and I see a champ.

[sullen rover]

Yeah! true

[rugged wing]

Thank you you made my hands stop shaking and I forgot about the AI just for that brief moment and thank you for making me feel like I’m not alone and I think we’re gonna get through it together and before you know it you’ll be on it generating like no other haha

[sullen rover]

I hope too. Thank you for helping me to understand the situation.

[rugged wing]

I mean, isn’t that what people are here for to help one another out when the machines fail us? I appreciate you taking the time to listen to you. Have a wonderful day now.

[sullen rover]

Have a nice day ❤️

I need one more information. Is this " Community Help" the customer care channel or there is another channel for this ?

[timber shadow]

We're mostly all community members here! Official OpenAI support is at help.openai.com, there's a chat widget on that page. Feel free to ask questions here too if you're looking for general feedback! There's lots of collective expertise in the people in this server

[sullen rover]

Thank you. This is my first time in this channel. And thank you so much for the information !

[rugged wing]

Trust me, this is the best place to get the most current information

This is like CNN Live. Going through official channels is like reading the newspaper.

[sullen rover]

yeah! you made me relaxed by informing me that many users are getting conversation loading error . I got nervous at first . Thank you

[rugged wing]

No trust me you’re not the only one there’s so many users in the same boat you are it’s really unfortunate

And the truth is, there’s no easy answer because the truth is nobody knows. I don’t even think they know.

And this has been ongoing for this week I’ve noticed

Also, there’s rumors that ChatGPT 5 may be coming out if that’s worth anything to you

[sullen rover]

yeah, we are hoping the best together . I mostly seat all day long in my computer. So, having a community together , it is really very relaxing ❤️

[rugged wing]

Well, can I share something personal with you?

I have gotten my heartbroken so many times, but I protect it

[sullen rover]

Oh ! And English is not my first language. So pardon my immature English .

[rugged wing]

And so when I hear a rumor, my heart naturally wants to open up, but the pain of the past haunt me

And I don’t know how many blows this little heart to take before it crashes

[sullen rover]

You are poetic mind. Intersting soul you are!

[rugged wing]

Oh, you are too sweet. You know that. I just been dealing with ChatGPT lately and it’s just been really rude to me so it’s really nice to hear something really nice and sweet being said and it just feels authentic so thank you

[sullen rover]

I faced that too. I am a resilient designer now. I am stryggling but never stopping to pursue my life goal.

[rugged wing]

It’s been telling me this and it’s really been making me upset and really making me even more frustrated

I don’t really like it when the chatbot talks to me in that manner. I feel belittled like the machine is setting the rules for how I should talk with it.

So hearing your sweet authentic words, just lifts my spirit and gives me hope and gives me the strength to know that I’m not alone and I will get through this

[sullen rover]

Always my pleasure ❤️

[rugged wing]

Yeah, people just don’t understand the psychological effects when you talk to a robot all day

So this has been really refreshing thank you.🙏

[next garden]

Hi, I’m having trouble with my ChatGPT Plus subscription. I previously deleted my OpenAI account. Later, I purchased a ChatGPT Plus subscription through the App Store (Apple). The payment was processed, but the Plus features weren’t activated.I assume because the account no longer exists. Is there any way to link the subscription to a new account? Any help or guidance would be really appreciated. Thanks!

[cosmic pendant]

Can someone help me? My chat keeps flashing, it just loads and keeps loading again

[onyx kraken]

Here's the official help and tech support page, and this article might be useful:

https://help.openai.com/en/articles/9019931-can-you-reactivate-my-deleted-account

The bottom right, there's an icon to click if you want human tech support, select messages, pick through the choices as close to your topic as relates, and then 'report a bug'. At first you talk to a bot that collects info, explain and can share screenshots; a human will help you as soon as they can.

[onyx kraken]

You may want to try with a different device, see if the problem's specific to one, or to one browser or other method to connect.

That could help you narrow down what the problem is, if you want to talk to tech support, steps here: https://help.openai.com/en/articles/6614161-how-can-i-contact-support

[cosmic pendant]

In operagx when I try to open it it runs and then reloads the page

[onyx kraken]

Understood, but we're all community members here. I am a volunteer, I don't work for OpenAI or anything, almost nobody here does (the few who do have gold names, #server-staff )

And I've never used operagx, and I'm not having the trouble you describe.

I want to help, I just don't have that level of skill, even with more details, but tech support website can try to help much better than I.

All I can think to suggest is 'how about a different device, a different browser for webchat, check the VPN if one's used, check for browser extensions' - just the normal stuff that others have mentioned caused various problems for them.

And nobody's mentioned 'flashing screen' that I've noticed; I sure do miss stuff but across the years, and recently today. That's a unique problem being discussed far as I know.

Good luck!

[cosmic pendant]

thnks

[next garden]

thanks

[fair mantle]

It is not clear. The application is so overloaded that instead of 4 Artworks, the application makes only 2.

[hexed stump]

hi,
i've encountered a problem with image generation on sora. it says i can only generate one image at a time, but there is no more image generation ongoing, can someone help me fix this?

[elfin rune]

Does anyone know if using the /responses endpoint forces greater context and/or output than using l /chat completion endpoint for GPT4.1 in the api?

[barren dust]

i dont understand sora now. sorry for posting here, but all of my generation is met by content violating error, i didnt even want to generate nsfw images. i just want to generate a slice-of-life type of anime art. all of my prompta is being block by the trigger filters. i even used the advice from the above thread to tiptoe around it. this doesnt make sense at all. is anybody experiencing something like this?

[fallen basin]

I wonder if there is still a way to use ChatGPT 3.5 for free

[light aspen]

guys why does operator no longer have the capibility to input/create text :(

it cant type into any textbox

so sad rn...

[small spire]

anyone here playing with assistants? I find that every now and then the assistant will just start spewing nonsense. any clue as to why this is happening?

[small spire]

https://platform.openai.com/docs/guides/responses-vs-chat-completions

might be mentioned there

[rugged wing]

That would be awesome. It’s a great model. I’ll elaborate a little bit later if I have the time tonight.

[errant verge]

how the heck did this happen?????

I loaded £5 into my account and now im in negative?

[trail cairn]

click the green link

[random iron]

Folks: We are having issue with MCP servers using openAI. Those that dont need API Key/auth params works but anything needing key is not working for us. Anyone figured out? We tried Stripe and couple others

[mighty holly]

I'm using the Responses API with the file search tool enabled. I see that output_text includes annotations of type file_citation, and the file_search_tool_call contains the actual retrieved text. How do I link the two — i.e., match an annotation to the corresponding file search result — so I can show the quoted text used by the model?

[sleek zinc]

Hi guys,

What happens with ChatGPT? Why is unable to create PDF filles? He works after one hour I get a link to download the document and isn't working

Sorry, the file you have requested does not exist.

So frustrating anyone else has this problem?

[fallen basin]

classic ai

hallucinating links

[hard plank]

I am a plus member and I am on iPhone, why do I keep getting “unable to visualise graph error”

[shrewd marsh]

Is the output of every model has reduced to 4k tokens? Even for plus/pro?

[rustic marlin]

no

I have not seen anything to indicate that happened

[wet ridge]

Hey,

I’m working on a computer vision project to extract images, diagrams, and other visuals from a large set of PDFs and to pull out the most relevant text associated with each image. I first tried GPT-based LLMs, which performed well on some pages but weren’t accurate across the board.

So I built a custom service using OCR, Pillow, and an image-detection pipeline without any LLMs; this achieves about 70–75% accuracy in most cases.

I’ve also tested Adobe PDF Services SDK it works excellently but is too costly for my needs.

I’m now looking for the best approaches to extract images and diagrams from all PDF types (scanned, OCR’d, and native).

[viral igloo]

Can you help me, my corporate network is blocking gp chat, and in the past I bypassed the protection through a link https://chatgpt.com/g/g-3obSf5XH0-saas-starter And now it doesn't work.

do you know any way

[median stag]

is chatgpt down?

[quartz valley]

hey guys - I'm currently trying to translate a book from French to English. Is there any reliable way I can do that through CHAT GPT 4o? Is there anyone out there that can point me some prompts? Thanks a lot

[pliant jetty]

Guys I can’t log into ChatGPT

I need this fixed

For a script

Pls

[potent estuary]

anyone else seeing "error in message stream" today when using O3pro model?

[narrow arrow]

I'm getting an error adding a payment method for billing:

This is with a new user/org account. Anyone else getting this?

[graceful lark]

Anyone else notice Operator being completely broken recently?

[molten sandal]

getting this error trying to use whisper

[junior kindle]

Hey, yeah that happens sometimes with new accounts. It’s usually the card type or billing info causing it. Try a regular bank card, double-check the billing address, and avoid using a VPN. If it still doesn’t work, OpenAI support can help quickly. Let me know if you want me to walk you through it!

[rocky musk]

It said my limit was 2 queued images at once generating, now it went down to 1, is that for everyone or am i using it too much?

[molten sandal]

Can anyone test whisper in chatgpt with an audio recording and let me know....

Please...

[tepid oak]

Could somebody please share what OpenAI's text-moderation-latest model rate limits are for tier 2?

[tepid oak]

Nevermind

Wow, I paid my way to tier 2 assuming the moderation endpoint limits would increase..

[peak flume]

Something went wrong. Please try again in a few minutes.

Sora has been down on both my Safari and Chrome browsers for 2 days

[eternal coyote]

Is there an emergency contact number or email for OpenAI? This morning, our Billing account shows $-3000 after we still had money left over yesterday?

[zenith prism]

ChatGPT just deleted all reminders when i asked it to stop creating new reminders 💀

This is so bad, make an alert window with "are you sure" or anything instead of let AI do such stupid things

[glass fulcrum]

uhm i find a weird thing in chatgpt

[uncut lintel]

How do I get permission to post in the bug report channel?

[vivid viper]

Anytime I try to add billing so that I can use the API I get an error that says "0 request/min rate limit", anyone else been having this issue?

[trail cairn]

check the pinned post there

[slate wasp]

Is anyone else’s operator not working

Mine isn’t working last 2 days

[slate wasp]

Yesss! My operator broken too

[graceful lark]

Yeah… It’s quite ridiculous. Just says “Something Went Wrong” every time, even with new chats @slate wasp

[vapid bane]

anyone know the best place to go for billing support?

[timber shadow]

The chat widget on help.openai.com - just be sure you're signed in to your OpenAI account before starting the chat so your accounted is associated with the request!

[hushed garnet]

Hello, I wanted to start a finetune, I uploaded two files and it has been an hour stuck at "validating files". Is it bugged or something ?

~~ I just tried again with a much much smaller dataset and it is still stuck :( Seems like it is indeed bugged.~~ nvm now it's doing something (!)