community-help

[south cairn]

down

[south cairn]

Yep just popped up upon refresh

[brittle moon]

Hi guys
I subbed to plus and got refunded immediately minus 2 bucks for fees.
I sent a bug report but they dont seem to reply

[onyx kraken]

https://help.openai.com/ often needs hours to days before someone can respond.

[brittle moon]

Then they shouldnt offer paid service imho.

[onyx kraken]

I've been happy with mine. Sorry whatever happened to you.

[brittle moon]

Its like talking to someone sick and telling them im healthy and fine sorry for you lol

[onyx kraken]

You just said they shouldn't offer paid service in your opinion; that's like telling everyone else they have no right to use or enjoy the service you choose not to keep.

Not very nice of you.

[zealous dagger]

Is chatgpt down for anyone else?

[onyx kraken]

Yep. https://status.openai.com/

[brittle moon]

I didnt choose not to keep, i got refunded automatically

[onyx kraken]

Sounds odd. If you reached out to https://help.openai.com/ and used the bot bottom right to 'report a bug' and left a message explaining your situation, then they'll help you when they can.

[brittle moon]

Alright thanks

[modest vapor]

Hi all, OpenAI Microscope has not been working for a while, any idea what is wrong?

[lucid coral]

Wanted to renew gpt plus because subscription expired (didn't have funds in a/c( now its not taking

[midnight patrol]

Hey can anyone help me with open ai evals?

[gray cosmos]

i am developing a chatbot using api and have a chat with chat gpt with all my history and code etc, i updated to gpt+ and now its giving me errors

anyone able to help

please

[real linden]

I had a conversation with GPT. Reaching the limit, I decided that I would buy a subscription. After the purchase, when I returned to the chat I am unable to continue the conversation because of the error shown in the screenshot. It appears every time I send a new message.
Refreshing the page, logging out and logging in again and incognito mode does not help. Do you have any ideas for this cause i need this chat?

[onyx kraken]

I heard someone else mention this last night, they also saw that happen right after they became a plus member.

It's possible that the previous conversation was somehow affected.

You should be able to start new chats, including copy/paste that entire conversation (chunk it if you need/want to, but pretty big inputs work too) into a new chat and tell the model you want to keep going from there.

[charred flint]

Hi OpenAI gurus. I have been using SearchGPT and find it very useful. I want to use the APIs. I don't see that they are available yet. Does anyone know how to get access to the Search GPT APIs? Is there a wait list?

[sturdy jewel]

The API is not available for search GPT, I don’t think there’s been any announcements about whether they’re even going to make it available

[charred flint]

bummer. Thanks for letting me know. If anyone has API based alternatives, please ping me. Much appreciated.

[eternal quartz]

trying to upload a image and i have this error (I have CHATGPT PLUS)

[tight valley]

Chat seems to be having issues again. anyone else having trouble getting anything submitted for an answer/

[onyx kraken]

All good for me

[idle hill]

hello?

anyone??

[onyx kraken]

Say what you came for, so folks know how to answer 😉

[idle hill]

chat GPT won't generate a high Intrinsic value IN high R^2 field with sound fluctuations

[onyx kraken]

Egads. Be sure to yourself provide examples of exactly what output you want from the model, taking into account the model's output limitations (should you want audio as audio, for example). You can invite it to use Python tool to help increase the range it can provide (like a graph showing this information).

[idle hill]

Coefficient of Determination

R^2

[idle hill]

so it wont help me?

[onyx kraken]

Give it examples of the output you want and it's very likely to follow those examples to the best it can!

Text is easy, sound is impossible (you can only input speech to it as sound, it won't get other noises like music or sirens or whatever) and some visuals can be worked around.

[idle hill]

i dont have examples text only

the output a is a R^2 high intrinsic field

never mind just realized i could google

examples

[autumn kestrel]

Hey,anything i would like to send to chatgpt this is the response:

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

I also have paid plan

i also tried to start a new converstation

[opal halo]

GPT is no longer able to edit my stuff on canvas for some reason, what can I do?

[arctic karma]

I have recived double charges for over a year. How can I get a human at openai to acknowledge this? Using the automated system, it refunded my one month and cancelled my account (assuming that I just wanted to cancel my sub). I need a person to see the issue but all help chats, emails etc are just bots.

[limpid sundial]

Hi

Hi

Hi

#welcome

Let

[mighty rivet]

I have 2 plus accounts and I get this whenever I am creating a GPT on one of them, the other one is fine. Anybody else?

[errant dock]

Let me ask what error is this? I have tried reloading the web many times but still cannot access chatgpt.

[onyx kraken]

https://help.openai.com/en/ bot bottom right. Select messages then pick through the choices until you can 'report a bug'. That gets you talking to a bot, it may ask for more details, and a human will respond in minutes to days, depending on how busy they are and how many techs are able to help your exact problem. Check back to that page and check the bot's messages to see the reply.

[rigid swift]

I'm getting the same page. Clearing cookies and cache have affected no change. Attempting to cheez the page by attempting to click 'login' before it changes to the 'oops' 'page is ineffective.

[zealous dagger]

is chatgpt down

[short patio]

not for me

[vale shell]

Is there anyway to have GPTs filling forms online automatically?
Ideally I would pass a description with some info and would get the online form automatically populated

[limpid ruin]

Lately the canvas feature has not been working for me at all. I switch to ChatGPT4o with canvas and ask it to open a canvas, and it doesn't know how to do it. Anyone experiencing similar issues?

[lyric ridge]

Update, my main OpenAI account just got reinstated, and I got an email from OpenAI saying so. Finally!!!

[haughty spindle]

how do i go to the ui from 1h ago

the new one is so annoying and unuserfriendly

[haughty spindle]

chatgpt is so unusable now i hate it

[crystal hemlock]

chat gpt bugged out and it wont let me send new messages, the stop button doesnt work either. reloading the page doesnt fix this issue...

[haughty spindle]

its been a bug for a while

im not sure what changed about it or when but ive experienced it a lot

[sturdy jewel]

if refresh doesn’t work, close the tab. if that doesn’t work, close the browser. if that doesn’t work, restart the computer… probably half of all IT problems right there 😁

[autumn torrent]

is chatgpt down again? it's never stable, even for paid accounts.

[idle hill]

i need help with .xml file save

the xml file deleted itself

[steel fox]

Hey! I need to prepare stats for my boss using two documents: one Word file (4 data points per entry) and one Excel sheet (8 columns per entry). The Word doc covers Aug 21 - Oct 21, and the Excel doc covers Oct 21 - today. Each has about 30 entries. Could ChatGPT reliably merge these and pull specific stats? Thanks!

[tiny turtle]

try with zoiper and custom GTP's!

[formal beacon]

I have a question, I use chatgpt to upload screenshots of my schoolbooks, so I can learn from its summarizes. Currently I'm using the free version and can upload a limited amount of images (I dont know exactly how many). If I get GPT Plus, how many images can I upload? Also, is it per chat, meaning that if I start a new chat I can upload again a certain amount of images, or is a total max. amount?

[warped relic]

Unlimited

[onyx kraken]

Yeah, so if you're a plus member, you get 80 messages with the full 4o model every 3 hours. You can upload images to that model with any of those messages, can upload more than one image in the same message, though the entire one chat has a limit, I think it's 10. But you can start a new chat and upload another 10 or however many.

You also get 5x as many tokens with the 4o-mini as you had before you became a plus member, and you sure can upload images to it as well. And there's other perks and options as a plus member as well.

[onyx kraken]

But yeah, there's no current stated limit on uploading images to the model, and I've never spotted a limit in my exploring.

[rare socket]

anybody know why, at certain times, chatgpt refuses to acknowledge uploads or use browser mode when instructed? (paid account)

[sleek hill]

The new Chatgpt windows app is broken and chats don't update and freezes, also how come we don't have the same feature where it's speech to text and when done recorcing we see the text then we submit, i prefer that

and when using the voice mode i should be able to interrupt Chatgpt so it doesnt yap

[half iris]

My chatgpt search doesn't work during the recent 2 days. I tried to tell gpt that please search the Internet then got an answer "I cannot search the Internet since I based on pretrained data"

Anyone else got the same issue?

[onyx kraken]

Works fine for me.

Are you a free member or a plus member? Are you asking in new chats?

Do you have any custom instructions set, or memories on? Either can potentially mess the model up if there's a conflict.

[half iris]

Oh I tried it on Android app and works fine just now. But still not working on web. No idea what happened hahha but thanks! Let me try more.

[onyx kraken]

If it's many refusals in one chat, that's probably just a bad chat. Once the model picks a weird part of its training data, it tends to keep answering the same weird way.

[half iris]

Starting a new chat couldn't solve it, still

[onyx kraken]

Interesting. Willing to share a chat where you try to search?

If you have memory and custom instructions on, if you have something in there that is triggering refusal, that can explain it even in a new chat. So I'd turn that off and see if that fixes the problem - if you can search without the memory/custom instructions on, but not with them on, then it's time to check through what the model has there and see if anything's about searching that might mess it up.

[half iris]

After cleaning the cache, web becomes alive again😆Thank you so much!

You gave me really awesome and nice help

[onyx kraken]

Thanks! I'm really glad you found a solution!

[rugged wing]

i’m using the plus version and I’m trying to generate an image, but it says it can’t do it. I also tried to search the web with the search function, but it says it can’t do that either. PS: it's happening only on my desktop

[idle hill]

Something went wrong. If this issue persists please contact us through our help center at help.openai.com.

[primal oracle]

Hey guys anyone having a tough time uploading PDFs today ? Please anyone tell me I am having issues with uploading PDFs

I tried using the Mac App, iOS App and the browser

[rugged wing]

???

[trail cairn]

what was the first message?

when you want it to look online you should not mention search or check online as it may confuse it to its training data where it had no such ability

[golden dove]

my gpt-4o1 not think hwo to fix?(

[uncut sable]

Hey, I was curious if I could get in contact with any researches working w\ open Ai.

[weak shell]

anyone get the new "work with" feature on macos to get access to Webstorm or any other IDE by chance? it's only letting me select terminal

[mortal turtle]

got it going with VS Code but only locally

[pine thorn]

is gpt broken?

Something went wrong while generating the response. If this issue persists please contact us through our help center at help.openai.com.

[burnt marsh]

which would be better to use as a research assistant for an academic research paper? perplexity or searchgpt (note: I only have the free plan for perplexity)

[haughty spindle]

that happens when u submit too many chats in an hour or so

[agile gale]

Is it broken right now? Whenever I regenerate a prompt, it says how can I help?

[agile gale]

Or freezes when I refresh

[stable canopy]

Hey! someone can help me with me? i have a subscription on my phone, but when i log in on pc i can't see the subscription. i logged in with the same method but on pc it's shows i have the free plan...

[rustic marlin]

are you sure it is the same account? does it show the same conversation history on both?

as a matter of fact, are you sure that the subscription that you purchased for your phone's app is actually ChatGPT and not a knockoff chatgpt? The app stores are littered with those

[stable canopy]

Its the official so that part okay

[stable canopy]

Well it should be, bc i signed in with google in both device, email address is the same... But on the pc the conversation history doesn't showing....

Which is wierd bc same email address and same method was used on both... 😦

[rustic marlin]

this seeems to check out

try to clear the browser cache and login again

[stable canopy]

I already did that, but still nothing 😦

[faint temple]

Is it normal that GPT 4o mini now has search with bing?

[north shadow]

Ok, seriously, at this point i am seriously mad and angry. This stupid Response 1 and 2 thing is seriously getting on my nerves. I did never ask for this stupid feature and do not want it. How the hell can i turn this nonsense off entirely? I hate it and do not want it. Does anyone know how i can ban this feature to the damn shadow realm? I do not want it at all.

[onyx kraken]

I don't think we can as is.

However #1070006151938314300 and request it and explain your reaction to the testing. I'm sure OpenAI is interested and cares.

[north shadow]

it feels like they dont care, this has been annoying me for months at this point. It is so damn annoying every time it comes up. I do not want to damn responses, i just want a single one, at the beginning i could right away just pick one of them, but now i cant even do anything till they both did load entirelly, wasting my damn time even more. It is verry annoying and i hate it.

[onyx kraken]

They can't read your mind 😦

It's not purposeful to annoy anyone, but it's a great idea to tell them you don't like it, so they can be helped to realize it may even upset some people.

We're all community members here, just like yourself. The very few of us who work for OpenAI have gold names, nearly nobody and I don't see any sign they read our chats here, this is for us to talk and help each other.

But #1070006151938314300 and #1070006915414900886 , they seem to read there and even answer. That is a way to talk to them, and I hope you choose to share so they can know. Otherwise they don't know and it's random if it ever changes.

[north shadow]

Well, my hopes are very low that they even care when a single random person from somewhere is annoyed about this. I did look online and i cant find anyone else that is annoyed by this, and i am very sure that they wont just do something to please a single person, so why even bother with it? Its hopeless and pointless anyway.

[onyx kraken]

Which is actually why someone like yourself making the suggestion can really help.

It's probably possible for them to add a way to opt out of preference testing.

But - they may not even realize someone (or maybe many people) would prefer to opt out of it.

If you and those like you who have a specific and personal preference that differs from others - the rest of the world may not even imagine your opinion exists.

That's not reason to feel offended, nobody can think of everything.

But it can be reason to speak up and make sure it's noticed that people like yourself exist, so things have a chance to improve.

[north shadow]

fine, i wrote something, i hope they actually care and see it, because this has been getting on my nerves for months now.

[onyx kraken]

I'm glad you spoke up! It probably would never change if nobody who handles it knew someone didn't enjoy it.

[north shadow]

yea, well, i still dont think they even care. I think they will never even look at it, thats how it has been with all of these big corporations or companies behind these things, and openAI is one of those as well. They dont care about the individuals using it, just about the big players paying them millions to use this. Otherwise they also would not just make teams be so that you Have to buy 2. I would totally upgrade to that plan if i could just use it for myselfe, i do not have the money to pay for that. They would offer something like that for single users if they would care about people like me.

[onyx kraken]

I'm sorry you feel that way. I feel cared for and helped, but that's just how my mind works. Even if I can't be a 'team' all by myself without paying for a crowd.

[north shadow]

well, im sorry, but not everyone has the money for that. I am very lucky i can even afford the Plus plan, i could afford it to pay like 25 or 300 max to have that team thing but as a like 1 person plan, but i cant afford to pay for 25 2 times and i dont even can use the second access, im all alone. It would be a total wast of money. I would gladly pay a bit more if i had better acces and stuff with the teams thing, like 5-10 more, but its far to much to pay 50 every month.

[rustic marlin]

why does it bother you so much? I usually just see it as getting a extra use for free tbh

[north shadow]

because like i said, i have to wait till it generated both replies fully until i can do anything, wasting my time. I only have so much time left on this earth, and even if its just a few seconds, it ads up over time. And i hate that.

[rustic marlin]

ohhh, I see

idk if that will work, but I think if you refresh the page, it will revert back to the normal display with reply 1 and reply 2 that you can switch between then

and also cancel it

[north shadow]

yea, well, thats just going around the problem, the problem is still there. A problem that they added and that was not there a few months ago. A problem they could fix very easy by simply giving us the option to disable this specific "Feature" Or removing it all toghether.

[rustic marlin]

that isn't a bug, it is a feature

all I can do is to suggest you a way to get around it

other than that, it sounds like something that one would be able to locally patch with tampermonkey

would be nice if they added a button to opt out of AB tests

but I think it is unlikely they would add that since the whole thing is built specifically to gather user feedback for training the AI

[onyx kraken]

Well, training data's optional. Maybe they'd make preference testing optional too. After all, the 'preference' of someone who does not agree to participate in preference testing, yet has no way to opt out, is more likely given as 'noise' instead of a meaningful preference, since they can't explain their actual option. Maybe someone like Ram would always pick the faster to generate reply, but I don't see consent and willingness to participate even that far.

[rustic marlin]

well, kinda, the opt out for model training is about not using the content of your chats as training data for the AI

but Im not sure if the feedbacks fall under that category

if you provide feedback about a output with the 👍 👎 buttons or pick a preffered one on an AB test, then I think it can be used regasrless of the opt out

[onyx kraken]

Nods I see it as a sort of variant of that information gathering from users. There's loads of reasons why someone might want to opt out of participating in chats for training, or in 'I prefer one answer to another'.

[rustic marlin]

it is an educated guess tho, I could be wrong about that

maybe if you opted out, the page could simple not be doing anything about the one you picked

and it would be pointless

but I work in programming with that sort of stuff enough to know that some times.. the front and back end developers simply don't have enough communication to allow something pointless like that situation to happen

Im not saying this is the case, just that it could be.. and Im not going to audit the UI's code to figure it out :V

[onyx kraken]

Ehh. "I prefer not to be asked" exists on the human range, as does "Oooh, ask me!". Over and under representation, by preference, is always a thing. But yeah. hearing 'someone wants to opt out' can matter and be doable too.

[north shadow]

this exactly shows my problem with this. You guys dont care about something like this, this "Feature" Does not bother you in the least. It bothers and annoys me to no end for months at this point, so much that my mood is completely ruined for the entire day when it comes up. This is something that is never going to be fixed, because its one of those things that 99% of people dont bother or dont care about and the devs because of that will never do anything about. If they would simply give us a on/off switch for these "Experimental Features" or whatever you call this one, that would fix it right away.

[onyx kraken]

Nah. You can't read my mind, and your perspective may also sway you towards presuming things about my mindset.

All my advice to you and discussion about this is about movement in the direction you say you'd like, if only it would happen; for you not to even get preference testing.

Yes, your viewpoint and preference is a new idea to me. Doesn't mean I don't value it or think it's something that should remain regardless of what you think.

[north shadow]

i also dont intend to insult or hurt with this statement, its just how it feels like to me, how it always was and happend to me with these things. Im just a single person, and the chance that something like this, a thing that bothers me immensly will get fixed is so small, its nearly pointless to even try. I just hope they fix it at some point, and i dont have to deal with it anymore. just wanted to make this clear here, i dont mean to insult or hurt with any of this, im just very frustrated and annoyed at this point and wanted to kinda vent about this thing. I put it into the suggestions channel but i dont think it will be seen, thats it.

[onyx kraken]

I hope the change you request happens. It makes sense to me, now that I what you share about it, that some don't like it, and worse, you can't opt out without fairly strong measures like finding another way to achieve your goals.

I like choice and options. I see only benefit in users being able to choose if they want to be part of preference choices or not. Hope the change happens.

[north shadow]

yea, me too, thanks for listening and your time.

[rustic marlin]

I completely agree

[onyx kraken]

I commented on your suggestion. Perhaps others will too over time.

[north shadow]

thanks

[final olive]

Try using an incognito browser window, or a different browser (chrome, firefox, etc). Sometime browser plugins can mess things up.

[viral gust]

Will there be any increase in the allotted advanced voice model use time for us paid users? It's really discouraging to start a session just to "1 minute remaining" and then I have to wait 24 hours for my next ~20-30 minutes.

Especially when I'm paying $20 a month.

[sturdy jewel]

Try setting a timer when you use it so that you know exactly when you’re gonna have one minute remaining

[shy kelp]

is there a version that is required for the CHAT GPT mac app

[charred parcel]

can you rephrase your question, a version of what?

[round hamlet]

i cant login to my chat GPT acc i used google to create it but don’t have access to it anymore i have the email but when resetting password i receive none

i bought plus so new acc isn’t a solution

no way i’m c0oked bc i login w google😭

[faint temple]

Can someone just please tell me why GPT 4o mini has search with bing

Proof

Is this a new feature?

[shy kelp]

When trying to download the app for Mac from OpenAI it says it’s not available for Mac but why I was thinking maybe it is Mac requirements for the app to work or something.

[quaint cobalt]

Hi. GPT-4 / GPT-4o can’t read files (PDF, TXT, ...). It says "Unfortunately, I can’t directly work with the files you send me or extract their contents. But you can copy the text here, and I’ll help you with searching and analyzing."

I tried it in GPT-4o mini and it easily read my PDF. I have a PLUS subscription, I saw on the Internet that there were problems with this before, even a direct instruction to use a code interpreter does not help me. Please tell me what can be done?

[woven palm]

Can anyone help me, my pc is running low on storage and i cant figure out why my apps using 66 GB of storage it's not showing in settings which app using this much storage, can anyone help me delete those stuff and figure out what's taking up so much space?

[final olive]

There are tools available that will search your drive and let you know where the space is being used up. From that, you may be able to figure out what app that is associated with.
For web browser specifically, some apps can us a lot of 'local' storage that may need to be cleaned up regularly. For Firefox, go to settings : Privacy and Security : Cookies and Site Data : Manage Data. Click 'Storage' to sort and see which sites are using a lot. I have found 16GB being used by an app there to 'speedup' local app response.

[charred parcel]

you need to have a Mac with Apple Silicon to use the Mac App

[worn quartz]

ChatGPT is vert slow since yesterday, i use it on Mac app.

[daring island]

anyone guys can tell if that diagram who i did are logical? and this should work?
In general whole aplication are heavily based on utils.py, where i want to put many classes who are responsible for various things, there will be 4 versions of that application (micro, mini, standard and full), where on micro and mini and i want to make classes without inheritance between them, but in standard version and full i want make inheritance between classes to avoid duplications of code
In general this scheme is for AiConverter - fully offline converter and merger for various types of files
/edit You can ask maybe why im asking here - since this converter is maded by 2 individuals - me and ChatGPT 😉

[sturdy jewel]

my advice would be get a minimum viable product working which converts just between a few formats, write some tests for it, and then revisit your grand design

[spare jungle]

I have yet to find a way to give feedback or complaints. please instruct me on how.

[onyx kraken]

There's several ways to give feedback or share concerns about ChatGPT or other OpenAI products.

https://openai.com/form/chat-model-feedback/ is probably the most detailed way and certainly the most private, ideal if you want to let only OpenAI know about some specific problem.

If you are content sharing in general, #1070006915414900886 and #1070006151938314300 work for those goals.

You can talk to OpenAI tech support through https://help.openai.com/ and the bot bottom right, select messages and click through until you can 'report a bug', that lets you type and upload images and you first talk to a bot then when possible a human tech support joins the conversation, that can take minutes to days depending on how busy they are and how many are trained to help your specific situation.

[daring island]

Yea, thanks, i want do it in that time, first ill try to make a micri version with much less dependencies from utils.py and later we see 🙂

[true yoke]

Anyone can help ,me out here..... can't figure out whaty's happening, cleared chache did all the jazz but i am confident the issue is from openai

Can't even sign in on the help page to send a message

[hexed dock]

hi guys, my usage tier is stuck on 1 even though i added more than 50 dollars and my first succesful transaction was in 2023 (but that was for 0 dollars somehow?). should i wait for my first nonzero transaction to be 7 days ago?

[onyx kraken]

The help page does have login help if you can't sign in there.

Bot bottom right.

[onyx kraken]

This is me not logged in, having clicked the bot's 'messages' then 'I can't log in' buttons. It's a 2-step 'now I can ask for help' path.

You do need to tell them who you are, thus their request for the email at the header

[true yoke]

thanks

[trail cairn]

document/file content size and chatgpt plan affects this a lot. no changes are done to small files but others will almost be affected due limits

[dapper wave]

regular chatgpt has the ability to export all of your your conversations. Does anyone know how you do this using the enterprise version? I dont see it anymore, (i am the admin also). Anyone see a solution to this maybe?

[indigo violet]

How do I connect ChatGPT for MacOS to WezTerm using Work With? I'm using the latest available version of both applications.

[rare linden]

chatgpt is acting up for me, i cant get a data export + i lost a conversation and it said it was unable to load it, i tried signing out and signing back in on mobile and it shows no recent chats

[pearl vigil]

same, just lost one of my convos i was working on

[onyx kraken]

How long ago did you request the export? They can take a few hours or a day maybe to be sent.

Do ask again if it's been that long or longer!

[rare linden]

ah, my bad

my main issue is though is chatgpt is broken for me right now, the status says nothing, im unable to load a conversation i was working on and i signed out and back in on mobile and my recent convos disappeared

[pearl vigil]

Can attest, was typing a prompt to a convo and it hit me with the conversation not found error at random, wont let me open the convo back up. doesnt show up on mobile either.

[onyx kraken]

https://status.openai.com/ looks good.

Can ask https://help.openai.com/ bot bottom right for tech support if your problems continue

[pearl vigil]

yeah ill skip on that, never produces anything. Thanks for the suggestion though.

[rugged wing]

o1 preview doesnt think anymore for me, why?

It just responds instantly

[hexed dock]

hi guys, my usage tier is stuck on 1 even though i added more than 50 dollars and my first succesful transaction was in 2023 (but that was for 0 dollars somehow?). should i wait for my first nonzero transaction to be 7 days ago? <- can anyone help me please? i need the realistic voice to get me through my divorce

[trail cairn]

wait. zero means your usage was so low that it was not worth it to do a charge and thus you never have made a payment

[hexed dock]

ok so probably wait 2 more days then

[jovial steppe]

unusual activity error for the last 96 hours...

5 different topics opened up on the help.openai.com. always not responding to my followups. annoying.

[hexed dock]

yes their support is nonexistent

[elder dome]

Guys i paid for chatgpt plus yest and still havent recevied the invoice nor the gpt plus have mailed them but no reply yet what to do ? My bank ac statement already shows paid to open ai

[teal briar]

Hi guys, I have a question.

Is it possible to fine-tune a GPT that I created in the "create gpt" section of the app ?

[timber shadow]

Hello! "Fine-tuning" refers to the process of training an existing, trained GPT on a specific set of data - this is something you can do as part of OpenAI's API services, but it's not something you can do on ChatGPT or with GPTs.

The screenshot you shared is the GPT Editor, as you know - the sections on the left, "Create" and "Configure", are both for editing the instructions of your GPT. This is different from fine-tuning - the instructions of a GPT are like a set of text that the GPT "reads" for every input/output it processes. You can do a lot with this, even if it's not fine-tuning! I believe we're allotted 8,000 characters for the instructions field.

The "Create" section gives you a chat UI that will then edit/write/rewrite the "Instructions" field in the "Configure" section for you. You're also free to skip "Create" completely and just edit "Configure" yourself directly. I recommend that if you do this, you don't switch back over to "Create", because the Create chat will overwrite what you've done in "Configure" sometimes. "Configure" is also where you can manage things like uploading knowledge files, enabling/disabling GPT abilities, etc.

[teal briar]

Okay, thank you for your answer !

[torpid palm]

Hello everyone, I just tried to put 5$ on my openai account to be able to use your api functions but it said there was not enough money on my card but then they still took the 5$ without me getting anything, does anyone know what you can do?

[torpid palm]

I haven't received any mail with anything either

[vagrant plover]

For billing support please reach out to https://help.openai.com

You may want to check https://platform.openai.com/settings/organization/billing/history just in case.

[torpid palm]

Okay

[rough path]

I hope this isn’t too broad but I have a project I have been putting together and need a little feedback. Wouldn’t a “hub” that can be used to allow either locally built, specialized models-regardless of platform, run a continuous fine tuning of a much larger, (such as OpenAI) model be possible using a few commands to have the smaller model generate data while the larger model is trained on the data, without any human input be a useful tool? And before I get too far ahead of myself does anything like this exist?

[sturdy jewel]

I usually think of smaller models being trained from the output of larger models, what sort of specific use cases are you thinking for this? There might be a way to achieve a similar end result using a different approach.

[obsidian brook]

hi all, i cant get the chatgpt app for mac to work properly anymore - responses seemingly get cut off in the middle. like its returning a response and then the app decides its done while it's still streaming and it never gives me a complete answer anymore. it also cancels all search and analyze requests. it works just fine logged out and logged in via browser, but i can't use the app anymore :(

i've done all the troubleshooting steps i've could - cleared all chats, cleared memory, logged in and out, permissions are correct. it only recently seems to be doing this and it seems like a serverside issue more than anything because it's so consistent in when the responses cancel. it's like the connection is dropping or something. it's so weird

app: 1.2024.310, on macos 15.1.1, and i'm the free plan

i'm gonna keep it uninstalled for now, but if there's any pointers on how i can fix chatgpt's functionality i'd be super grateful

[onyx kraken]

Might want to contact tech support, https://help.openai.com/ bot bottom right, messages and pick through choices until you can 'report a bug'.

We're community members here like yourself.

[obsidian brook]

yeah i understand, was just seeing if anyone else was running into the same thing. will report though, thank you

[obsidian brook]

i figured it out, it was my vpn

it would be nice if the app returned a network error like the web app does

[arctic holly]

What did they do to the design they made a collapsible sidebar mandatory and it made it harder to find your old conversations in is no longer in chronological order from when you last used it. You have to now go searching like Windows 8 to make it worse you can’t see the stores of custom GPT‘s. They turned ChatGPT into Windows 8. What happened?

Also gpt stores and customs disappeared

Why did they force this new change on everyone ? This is ridiculous

I know MS has a close relationship with OpenAI but did MS send over their H1-B Visas to Windows 8/11 the design ?

[arctic holly]

Oh this is the canvas update and your forced to use the new style whether it’s on canvas mode or not jeez it sucks

[tame rune]

what should i do

[orchid rapids]

Is it possible to submit a PDF via the API (or a URL for a pdf) so that OpenAI can analyse it?

which models currently support this?

[orchid rapids]

Building an application with OpenAI api that analyses multiple PDFs with bank account statements. What's the best way of doing it?

I have multiple bank accounts in a few different countries. I want to be able to ask questions about it.

HOW I CURRENTLY DO IT:

1. I download all of my bank account statements (PDFs, CSVs, images...) and my family's (~20 statements, some are as long as 70 pages, some are 2 pages).
2. I upload them to ChatGPT.
3. I ask questions about them.

THE APP I WANT TO BUILD:

1. I upload all of my bank account statements to the app .
2. The answers to a set of pre-defined question are retrieved automatically.

HOW DO I ACHIEVE THIS?
I'm new to using the OpenAI api. I don't know how to achieve this. Some questions:

1. Can I submit PDFs, CSVs and images all through the same api call?
2. Which model can do this?
3. For the specific case of PDFs: is it better to
   ....a) convert to image and have openai answer questions about images? or
   ....b) extract text from the PDF and have openai find answers to questions on text?
4. Are there going to be problems with very long PDFs? What are some techniques to avoid such problems?

[rapid egret]

Hi, does anyone know why the retry button is gone when using Custom GPTs?

[near schooner]

Hello, I was talking to chatgpt about my research paper however I wasn't logged into my account since I was on a different device. does anyone think they could pull up that conversation? The research paper was about AI and robots replacing jobs in the workforce

[normal cedar]

I have built something similar (for a different use case) using make dot com which has a connector with the API, in my case we have all the files in one google drive, Make dot com loops through it and puts it into a vector store. It may work for you.

Make is quite a powerful low code no code way of orchestrating things.

When you then have it working them you can then go from there.

[south hemlock]

hey i just read about prompt caching which sound great for our use case since our prompt has a large static part. the docs say that it is automatically enabled for prompts that are 1024 tokens or longer. but our prompt is only 636 tokens. can we somehow enable it still?

[thin prawn]

I need a document solution..Can anyone please help.

[harsh steppe]

I recommend making a wish to the Flying Spaghetti Monster or Joe Pesci. It will be just as effective as any other method out there.

Even the dude who lost his bitcoin wallet in the dump has a better chance of finding it than that.

Here's an air hug that I hope makes you feel a bit better. air hug

[whole helm]

hello, i have this problem

i try the majority of recommendation but i can´t access whit my google account in firefox and edge

sorry the broken inglish

[onyx kraken]

While I don't know why it's gone, I know how you can work around.

When you want to regen 'this most recent output', instead of entering a new input, scroll up to your last input. Edit it, then submit, even without changes (changes are fine). If you don't change anything, the effect is the same as regenning the model's answer, except you scroll through your multiple inputs instead of its multiple outputs if you want to move between them

[onyx kraken]

That isn't possible.

'free no-account convos', which is what you have if you are not signed in, are not saved at all.

Best you can do is talk to the model again when you want (logged in or not) and redo the conversation or give it new instructions that you may like even better.

[rapid egret]

Thanks! This might be the only way for now

[wicked igloo]

please can someone help me? ive not deleted my account nor deactivated it. I cant get hold of anyone at OpenAI

[rustic marlin]

this is something that you will only be able to get sorted out with the support at the help center https://help.openai.com

[orchid rapids]

"puts it into a vector store": can you tell me more about it? what's the purpose of this and how does it work? I'm a noob

[rustic marlin]

vector databases is a way to structure data which relies on the way AI does the semantic embedding of data, allowing you to search the database by "semantic proximity" (oversimplification but that is the gist of it)
there are many open source projects for that sort of dataabse that work well with already existing databases and with libs for python, JS and other common languages, just search the web and yo uwill find plenty

it can be a very nice solution to some indexing and querying challenges with broader scope

[blazing rover]

hi guys, I need some help 😄
I use ChatGPT for image generation because it does pretty well at making solid, default characters. Then I also use img2img to expand on some of the characters and it does quite well at that too. There's one issue I'm having though where:

1. If I want to keep track of my images and expand on them with their generation IDs, my chat window starts to lag a lot. Is there any way to fix that? Would using the API outside of the website somehow help?
2. It is a bit slow and I can only generate 1 image at a time. Any way to do that better too?

[vagrant geyser]

I think I might need some help. 😂 Since I verified my phone number when registering for ChatGPT, but this phone number is no longer mine, I realized after completing the subscription payment that it might cause some issues if payment-related messages are also sent to the phone. Therefore, I’d like to ask if I subscribe to ChatGP Plus, will the phone number associated with my account receive SMS notifications?

[brave walrus]

https://help.openai.com/en/ I would contact them directly. You can go here and click on the little button on the bottom right. The little speech bubble and go through there to contact support.

[vagrant geyser]

Thank you, mate!

[normal cedar]

"puts it into a vector store": can you

[silver rivet]

could anyone give me a hand with this {"message":"Failed to evaluate expression '={\n "errorMessage": "Failed to process diagnosis: " + error.message,\n "errorCode": error.code\n}\n' in context '{"errorMessage":"Failed to process diagnosis: Too Many Requests","errorCode":"429"}'. Reason: Cannot parse '{"errorMessage":"Failed to process diagnosis: Too Many Requests","errorCode":"429"}' as 'io.camunda.connector.runtime.core.error.ConnectorError'.","type":"io.camunda.connector.feel.FeelEngineWrapperException"}

[brave walrus]

Reads like a rate limiter got hit. Ask Robin.

[hardy jasper]

I have the Plus version of ChatGPT, but since I upgraded 2 months ago, the microphone icon has disappeared from the website, so I can no longer use the microphone to dictate my questions as I do on my phone. I use ChatGPT on Google Chrome, and it's fully updated. However, there's no way to use the microphone on ChatGPT as I could with the free version. Does anyone know how to fix it? tnx

[sturdy jewel]

what browser are you using? Have you tried disabling extensions or plug-ins and restarting the browser? You may need to clear the cache

[shell solstice]

I upgraded to ChatGPT Plus with a different email. I used the service but today I cannot login or receive a password reset to my new email address.

[harsh steppe]

why not update your account with the new email address?

[timber shadow]

Hello @agile tree just replying to your post in #off-topic: have you tried signing in fresh in a private window of Firefox yet, as a test?

[agile tree]

Just tried it, it works fine in private window, so the issue is just in a regular window?

[timber shadow]

Hmm then you might try clearing your browser cache, then closing/reopening it. Sometimes that just needs to be cleared out, if you haven't already tried it.

[agile tree]

Already tried it, still having the same issue

Only difference it shows the initial text, but no response to any prompt

Can't even logout or see past conversations

[final olive]

If works fine in private window, but not in regular window after clearing cache, then a place to look is browser plugins interfering in some way.

[timber shadow]

Ah dang ok. The other reasons why it might work in private but not regular are cookies and extensions. So, if you feel like it, maybe try disabling extensions and clearing cookies, clearing cache again, and restarting again.

[harsh steppe]

worst case, just uninstall and reinstall firefox lol

[agile tree]

I started firefox with plugins disabled runnning firefox --safe-mode but still didn't work

[timber shadow]

Could there be some browser setting enabled in regular sessions but disabled for private (or vice versa)? Some kind of privacy/cookies settings, perhaps?

[final olive]

A less aggressive path than full browser reinstall, is to use a profile. I do that to keep some things separate. Though I would have expected safe mode to have gotten there too.

[agile tree]

Hmmm I just reeboted the browser to default settings, looks like it solved the issue, not sure what happened at all

[harsh steppe]

elves, it's always those dang Keebler elves.

[agile tree]

Thanks for your help!

[red oyster]

Unable to Submit Messages in Chat Interface (Send Button Missing)

Hi everyone,

I'm having trouble with the chat interface in one of my ongoing conversations. Specifically, the "send" button (the up arrow) is missing, and I can’t submit any messages.

I’ve already tried a few troubleshooting steps, including refreshing the page, clearing cache and cookies, and switching browsers, but the problem persists. This issue is preventing me from continuing an important conversation.

Has anyone else experienced this? Are there any additional steps I can take to resolve this?

I’d appreciate any advice or assistance. Thanks in advance!

Best,
Yvonne

[quartz spruce]

hey my friends i have a large python app building and also a large zshrc both are over 18000 characters or around 2000 words give or take. i cannot for the life of me get chatgtp or custom gtps to write me anything above or around that many words or characters please any advice. its like it always wants to rush it. but i cant run my code because i keep getting bits and pieces and it "pretends to forget or its not important to add" thanks guys.

[sturdy jewel]

Break it into pieces, do it one piece at a time

[onyx kraken]

You can have the model help you understand how to modularize and chunk the code into pieces it and you can comfortably work with

[blazing rover]

Hi guys, is it possible for ChatGPT to make images in bulk and keep the same quality?

[sturdy jewel]

Congratulations, you have advanced to where you have to start learning about the API 😎 great things are ahead https://platform.openai.com/docs/overview

[onyx kraken]

I love how you make that sound!

remains a coward that keeps hidin' from that advancement

[sturdy jewel]

For true, me too, I’ve only played with it a bit. I tend to just read the docs and look at examples for fun.

[blazing rover]

😂 I can do it through the API?

cool

[weak ruin]

Why is Openai Support not refunding my money

I added to the API

[brave walrus]

I think the answer is actually just simply, "No, they need to be separate API requests." but I could be wrong.

[brave walrus]

I'm 50% wrong. sequential tool use is totally possible in chatgpt, just has to be specified. The restrictions for the number per API call, not total number of API calls per chat msg...

[onyx kraken]

I commonly ask the model to generate 5 images in the same prompt, made 1 at a time.

[dense cove]

blue dot means a message has new things (probably from mobile use) but everything up to alst 30 days has nothing? how do i get rid of the blue dot lol

[half salmon]

why I get that error
Could not fetch usage for this organization, if this issue persists please contact us through our help center at https://help.openai.com.

[quartz spruce]

how would i word this bro?

[onyx kraken]

Adjust this for your actual project, but I'd ask the model something like,

"Hey, this is huge. Too long for one discussion.

How do we break this into chunks that we can work on and not get lost or overwhelmed? Where's it naturally split into components?

[quartz spruce]

Thanks man as always

[sweet pulsar]

isn't chatgpt 5 close to release?

[regal lynx]

No update on a new model yet. All official updates or releases will be posted to #announcements

[sweet pulsar]

kk

[cerulean blaze]

is there anyway to change font size on the chatgpt windows app

[trail cairn]

CTRL + and -

[inland pebble]

Quotes from PDFs not visible on custom GPT answers… how can I fix this problem? Please help

[fresh flume]

Hi, does ChatGPT still commit to memory and it's doing so discreetly? I haven't seen the notification about it saving anything to memory for I think at least a couple of weeks.

[onyx kraken]

Commits to memory are still visible.

The only two I've seen in a couple weeks I had to ask for, but my memory is at 96% and I figured the model was being careful with the remaining space and/or everything's already covered.

[onyx kraken]

Since the model chooses when to use the tool, I checked. How it uses it sure changed, it used to remember when I mentioned products that I owned; when I asked why that it told me then so it could offer me tailored advice related to my stuff I have.

Here I tell it I love my new brand car and ask for maintenance tips; it gives them but doesn't make the memory about the type of car. That's a huge change from before.

I ask about what triggers memories and for examples.

It gives them, but might be making them up; maybe for now at least some of us have to explicitly tell the model what we want remembered.

This could be related to how full the memory already is; the model might make choices based on how much it already knows about the user and if that's 'enough' to personalize the conversation or not.

It also couldbe that the model only makes memories right now when told to (but I still see posts from people asking how to get it to stop remembering every little non-important detail - so I do wonder if the model is trained to add memories until it has a profile of 'some level of quality and usefulness' and then be more sparing.

https://chatgpt.com/share/6744b0b8-6aac-8011-a1c9-28cb0d1851b6

[fresh flume]

Thanks for the response. Mine was also full, which is about the time I stopped seeing the notifications. I went in and deleted a bunch from memory but still haven't seen them, so it's good to see that it changed and it's not just me at least.

[slender shell]

whats going on with chat search, has it been disabled?

[timber shadow]

Hey, do you mean SearchGPT? If so, it hasn't been disabled, it's just been moved. It's the new button next to the attach file button in the bottom-left of the new message box.

[brave walrus]

search.. like eh, this? or...

or "search" like this?

[timber shadow]

ooo yeah I forgot about that chat search. fwiw that only appears in the sidebar - not visible if sidebar is closed!

[slender shell]

Was looking for the search chats button, i think it must've glitched. its back now thanks

[tardy igloo]

Ho

[full lynx]

So, I have this problem where for some reason ChatGPT has lost access to the browser tool when I ask for it to search for something in our regular chat. Not only that but it for some reason won't remember things we just talked about or distorts the memories it has saved when I ask it to repeat some of them? This has happened a few times now and I was wondering if anyone knew what the reason for this would be.

When I ask it to search the web it always stuff like "I don't have the ability," and "My knowledge is up to date until my last training cut-off" but it was just fine a few hours ago?

[sturdy jewel]

Do you still see the same behavior if you open a new chat window?

[full lynx]

Currently I've tried refreshing, logging out and back in, and opening a new chat window. So far it's stayed the same

This happened like a week ago as well but it fixed itself after a couple of hours. This time it's been going on for over 12 hours and counting.

[timber shadow]

For the browser tool: just double checking that this box is checked in your Customize ChatGPT section. For the rest of the stuff, you said you tried opening a new chat window, I just want to double check that by that you mean you have bugs like this in both existing and brand new chats?

Also, do you have this button in new 4o chats/does it work?

[full lynx]

A picture of it's response. For the record, I've only used the regular ChatGPT. I do have that button but it doesn't work currently even when I try it

[timber shadow]

Have you already tried a brand-new chat, something like "Search online for new tech news"? Wondering whether it's all new chats too, or just this one chat.

(fwiw that phrasing, "Why can't you..." could confuse the model - if it takes you at your word that it already "can't", then it could be more likely to hallucinate that it cant)

[full lynx]

Brand new chat

[timber shadow]

I just tried a couple searches on my end too and it bugged as well! Maybe a hiccup related to the other hiccup earlier today

https://status.openai.com/

[full lynx]

This may sound bad, but I'm glad it isn't just me lol. I thought something I did like overloaded it or something I was genuinely worried.

Thanks for taking the time to respond to me

[onyx kraken]

So, not sure how much is related to custom instructions and other ways to get the model to 'think' about what to do.

But my typical RP vague request to search worked fine right now. Appears to successfully search.

[timber shadow]

Nice, search seems to be back up for me now too!

[onyx kraken]

Been an hour since you had that trouble, if it was the system maybe it's clear now. I can use your prompt okay.

If you have custom instructions on, and/or memories. There is a chance you have some instruction there that conflicts in some way and guides it away from searching.

This is less likely if the problem comes and goes across long periods - if it's your instruction it should work the same most of the time, randomly failing if it focuses on the conflict, randomly succeeding when it focuses on parts that don't conflict. Chunks of clumped error or success suggest a possible account issue if it's only affecting some of us (I don't seem affected).

I recommend tech support, https://help.openai.com/ bot bottom right. Select messages, pick through the choices closest to your situation until you can 'report a bug'. Let them know about it; can take minutes to days to get a reply. But you could need to have something about your account fixed if this has been happening on and off for a while. Check that page and the message in the bot periodically to see when a human's answered.

[full lynx]

When I tried again. I'll try tech support to see if I can get some assistance there

[cerulean blaze]

dosnt work on the desktop app

[brave walrus]

?

[cerulean blaze]

wait when did this get added becs it never used to

[brave walrus]

In any case. Seems to be working now! Enjoy.

[cerulean blaze]

is there a shortcut in window cut to load straight into voice mode i wanna add it to my streamdeck i know there no plugin so i can set a button as a short cut to open a app if i can make a short cut with the "--" part i can do it if not then the button will load up chatgpt and not voice mode

[unreal talon]

Hello all, there seems to be some issues with my chatgpt plus account. When I tried to ask him to create a photo and use website search, it didn't seem to work. In addition, the thinking phase of my O1 series models has disappeared. What is the situation? It seems that AI has become stupid(；′⌒`)

[timber shadow]

#ai-discussions is the best place for non-OpenAI stuff

[rain zephyr]

Oops. Sorry.

[junior warren]

Is there a moderator available?

Please tag me.

[timber shadow]

Heya! You can reach out to the server mods by going to this profile: @burnt galleon, then tap/click the message option, then send a DM. When prompted, select the option to send the message to the server mods 🙂

[junior warren]

@timber shadow thank you very much.

[junior warren]

Hopefully the second part of my request was sent through as well.

Thank you.

@timber shadow

[rotund holly]

Hello! I’m sorry if this wasn’t the right channel to ask this in but I need help 😭 so I won’t lie I use open ai for janitor ai and stuff but I keep getting this???

And at first I thought I used up the monies I had there but I checked and I didn’t, I have money on there and in my usage limit, I set the usage limits to “none” so I’m just confused???? Is this an OpenAI thing or is something wrong with Janitor?

[trail cairn]

what does it show here https://platform.openai.com/settings/organization/billing/overview

[rotund holly]

Let me see!

[rotund holly]

It just shows me this

[trail cairn]

if you bought the credits within one hour you might need to wait a bit longer

also note that openai has history of banning accounts that use the api services for janitor

[rotund holly]

Ohhh okay! I’ll just wait a bit longer but it has been a while that last time I bought credits was around 2am this morning and it’s 4pm now

[rotund holly]

OH

no I hope that did not happen to meeee 😭😭😭😭

[onyx kraken]

Apparently, that service is claimed to itself have, at least at one time, added instructions as a pre-prompt to your chat through it, so stuff that breaks https://openai.com/policies/row-terms-of-use/ and/or https://openai.com/policies/usage-policies/ was sent with whatever you input; and so everyone going through that service to interact with the OpenAI model was affected when OAI took actions about it, because we're responsible for what we're sending per ToS.

https://community.openai.com/t/janitor-ai-account-terminated/281969/17?page=2

As that link suggests, reach out to the other service maybe and discuss with them; good luck!

[delicate goblet]

does anyone know why im being forced to put in a verrification code from my email when i have 2fa enabled with an authenticator

it never needed it before and its really annoying

[timber shadow]

Heya, this channel is not private, so you should delete the pic with the email in it 🙂

As for what's happening: is it doing email auth instead of auth app, or is it doing both?

[delicate goblet]

both

[timber shadow]

So you enter email + password, then email auth, then auth app, in that order? Or do you not sign in using email + password?

[delicate goblet]

email + password, then email auth, then auth app

it just started today asking for the email auth

[timber shadow]

What do you see at https://chatgpt.com/#settings/Security in the MFA section?

[timber shadow]

Strange! I wonder if it might be just a temporary bug? If you feel like going through the trouble, maybe try disabling/reenabling MFA to see if that clears it up. I wouldn't imagine it has to do with any suspicious account access but if you have any doubt it can't hurt to change your password too (ideally after cycling MFA, if you do that)

Do you have to sign in frequently?

[delicate goblet]

already tried disabling/reenabling MFA

[delicate goblet]

yea i use multiple computers and i have the authenticator synced with bitwarden

so it just instant sign in usually

im guessing its a temporary bug as well

its fine for now i just hope it gets fixed

[digital plover]

Is there a way to pin your favorite chatgpt conversations?

[quasi forge]

I don't think so, but you could archive them

Is there a way to block your custom gpt from accessing a specific site? My GPT keeps using the Fandom version of a game's wiki instead of the official one despite prompting to never use or reference it

[onyx kraken]

Have you tried telling it what to do? Don't even mention what not to do, discuss in detail what you do want.

[quasi forge]

That was how it worked originally, before I tried telling it to avoid Fandom. I'm testing it again but its links aren't working and its sources page is blank, maybe a bug with ChatGPT itself

[sturdy jewel]

I’m not sure if prompt instructions, or custom instructions—or memory for that matter—can influence the way the search tool works

or if they can, it’s more limited and not as targeted as saying use or avoid a certain site or type of site

[astral iris]

Hi everyone, I just joined here and looking for some help. Have you encountered the same problem as me, sometimes I can't evoke the canvas in the "4o with canvas" model, no matter what prompt I use, it will be fine after a while, and then it will fail again. Is it because I have hit the usage limit, or I violated the usage guidelines and was restricted?

[hasty salmon]

what happened with my gpt? the gpt refused to read the document
it said:
It seems there was a misunderstanding. I am unable to access or view any files uploaded, regardless of whether you're using a premium account or not. I can only provide assistance based on the information you share in text form.

Perhaps in a previous interaction, you shared a summary or key points from your paper, and I helped craft questions or discussions based on that.

If you can provide details or a summary of your paper, I'd be happy to help you create complex questions and responses based on it.

[orchid rapids]

Is OpenAI o1-preview being lazy? Why is it truncating my output?

I'm passing the o1 model a prompt to list all transactions in a markdown. I'm asking it to extract all transactions, but it is truncating the output like this:

-     {"id": 54, "amount": 180.00, "type": "out", "balance": 6224.81, "date": "2023-07-30"},
-     {"id": 55, "amount": 6.80, "type": "out", "balance": 5745.72, "date": "2023-05-27"},
-     {"id": 56, "amount": 3.90, "type": "out", "balance": 2556.99, "date": "2023-05-30"}
-     // ... (additional transactions would continue here)”

Why?

I'm using tiktoken to count the tokens, and they are no where the limit:

    encoding = tiktoken.encoding_for_model("o1-preview")
    input_tokens = encoding.encode(prompt)
    output = response0.choices[0].message.content
    output_tokens = encoding.encode(output)
    print(f"Number of INPUT tokens: {len(input_tokens)}. MAX: ?") #  24978.
    print(f"Number of OUTPUT tokens: {len(output_tokens)}. MAX: 32,768") # 2937.
    print(f"Number of TOTAL TOKENS used: {len(input_tokens + output_tokens)}. MAX: 128,000") # 27915.

    Number of INPUT tokens: 24978. MAX: ?
    Number of OUTPUT tokens: 2937. MAX: 32,768
    Number of TOTAL TOKENS used: 27915. MAX: 128,000

Finally, this is the prompt I'm using:

prompt = f"""
Instructions:
- You will receive a markdown document extracted from a bank account statement PDF.
- Analyze each transaction to determine the amount of money that was deposited or withdrawn.
- Provide a JSON formatted list of all transactions as shown in the example below:
 {{
     "transactions_list": [
         {{"id": 1, "amount": 1806.15, "type": "in", "balance": 2151.25, "date": "2021-07-16"}},
         {{"id": 2, "amount": 415.18, "type": "out", "balance": 1736.07, "date": "2021-07-17"}}
      ]
  }}

Markdown of bank account statement:###\n
{OCR_markdown}###
"""

[sturdy jewel]

In my opinion, you would be better off using ChatGPT to write a python script that extracts what you want, rather than trying to have the model itself do it

[orchid rapids]

interesting take. would that follow the flow 1. submit PDF to OpenAI (or submit markdown?) 2. ask OpenAI it to create a script. 3. ask OpenAI to execute the script and return the values OR how do I go aobut the execution part, because the model might truncate the output anyway.

In my case this is not a one off, I'm using MULTIPLE bank account formats that I don't know what their formats are beforehand!

[frigid swan]

Hello, do anyone know why i have a limit of 15 minutes/month for using the advanced voice funcitonnality on free plan, but my brother doesn't seem to have any limit on his account (he is also in free plan) ?

[severe quail]

Is it possible to have a chat that's too long? Chatgpt said it's about a 100-150 pages and a couple of days ago I've been getting something seems to have gone wrong every time I make a input into that chat

[regal lynx]

I recommend opening a new chat and seeing if that stops the issue. I've experienced some degradation in how fast I get responses with really long chats

[severe quail]

Opening as in continuing the story in a new chat?

[severe quail]

It does generate a answer in the chat but when I close the app it disappears and I have the error message, other chats are just fine

[regal lynx]

I recommend using a new chat then. My hope is that will fix the issue

[severe quail]

Probably, but the story doesn't feel the same tbh

It's pretty much the same issue as this https://discord.com/channels/974519864045756446/1290382310860132426

[sturdy jewel]

I was thinking more like write the python script with ChatGPT and run it locally, you could upload a small data set for testing in the app, have it write some pytest tests too (or unittest ones)

[minor pike]

hello

wwho else can help me

wwho know how to solve it

[sturdy jewel]

We try to encourage people not to post the same question in multiple places, so that folks aren’t duplicating effort trying to help people.

I think anyone trying to help you would need more information, if you could explain what you’re trying to do, maybe someone can help

[minor pike]

Oh, I know. Sorrry

so I ask my chat gpt it's just only respond like that. Could you help me

[sturdy jewel]

I understand you’re trying to get it to explain something, but it’s not clear what you wanted to explain if you just say “do it” do what? Help you with what?

[minor pike]

like thiss

[minor pike]

i send my imagination

photo

My english is not doood

good

[sturdy jewel]

Maybe start a new chat, upload the photo, and say in the same prompt “please explain what’s in this photo in English and Vietnamese”

[minor pike]

yes I mean Chatgpt do it and explain it for me

[uncut aurora]

You are using a GPT, do you? Is there a reason for this?
And what is your goal? Did you attach an image or not?

[minor pike]

yes I have sent it

[minor pike]

yes I really send all your requires

[sturdy jewel]

What model do you have selected?

[minor pike]

4o bro

[runic fiber]

hey guys it looks my Chatgpt does not open. Is there a down or smth?

[minor pike]

i think this man have same problem

[uncut aurora]

Please try it with GPT-4o mini and GPT-4.

[runic fiber]

I have no history chat , no allowed to upload file ... It's kind of blocked and I would like to log out but no options are there

[uncut aurora]

There is nothing recorded yet.
-> status.openai.com

[runic fiber]

Ok I managed to log out and in again, now it's working.

[minor pike]

so sometimes I asked and chat gpt only answer: Got it,.... feel free

so how can I fix it

[sturdy jewel]

A few things to try 1. Log out and log back in. 2. Try another web browser. 3. Restart your device.

[minor pike]

tks bro

[faint yacht]

Why can I no longer edit my own gpt in dialog mode? There is only one button where there were 2 buttons yesterday?!

[timber shadow]

Interesting, I'm seeing the same thing on my end! I wonder if it's a temp bug or what? You might already know this, but fwiw, the "Create" chat only writes the Instructions field in the "Configure" tab for you, so you can edit the Instructions directly and achieve similar results re: updating your GPT's behavior!

[faint yacht]

I will wait a bit 🙂 If this is a bug, perhaps I can work tomorrow on this 🙂

[severe quail]

Can I chat actually be too long? Or has my AI just given up

[timber shadow]

In a couple ways, yes! The most basic one is the context limit of the model itself. All LLMs have a limit to how much input they can take - think of it like a pipe. Once you fill up the pipe from one end, stuff has to leave the other end once it's full. The context limit of an LLM is exactly like that - it can't remember stuff indefinitely, and it will only be able to "remember" the most recent context up to its context limit, which varies from model-to-model.

Otherwise, on ChatGPT, chats have been known to sometimes get buggy or to stop working altogether once they reach very long lengths. In my experience, this has gotten more rare, but it sometimes happens! Usually starting a new chat (copy/paste over relevant context as needed!) gets things going again quickly.

[sturdy jewel]

Almost every time I’ve had problems with ChatGPT, it was fixed by opening a new chat

whenever a conversation is going well, I start to copy stuff out of it for a document

[royal scroll]

To be fair it remembers better than humans

It’s just worse at managing it

I can prove it

[severe quail]

Well according to chagpt my chat is about 100 to 150 pages long, and it just randomly started to give me error messages

[timber shadow]

LLMs don't work in terms of pages, they operate in units called tokens. LLM context limits are thousands of tokens in size, and 75 words is usually around 100 tokens (https://help.openai.com/en/articles/4936856-what-are-tokens-and-how-to-count-them). You might be running into the context limit of the model you're using in the chat!

[severe quail]

So apparently it has around 153,831 words

[timber shadow]

You could probably copy/paste the chat into a word processor and check!

[severe quail]

I pasted it into word

Should still work

[severe quail]

So its 203,585 tokens that that chat has

[timber shadow]

Nice! So depending on what version you're using, you're probably working with either an 8K or 32K context window:

[severe quail]

Im using the free version rn

[timber shadow]

So that means, every new message you send, the model is only "taking in" roughly the most recent 8K tokens

[severe quail]

So would the plus actually help? Or is the chat just gone for the time being

[timber shadow]

Plus has a context window 4x higher than on Free, it's a 32K context window. I don't know if a chat you started on a Free account can be continued with 32K context if you sign up for Plus, I've never tried that! But all new chats on Plus will use the 32K context window.

[severe quail]

Well hopefully

Chatgpt says i can still use the chat

[timber shadow]

One thing to keep in mind: ChatGPT can't deliver perfect answers about itself. It also can't count. So any questions you ask like this are likely to give hallucinated answers in response

[severe quail]

True

[timber shadow]

You can almost definitely still use the chat, I'm just not sure if it'll "switch over" the model used to the 32K version of 4o on Plus - but honestly I'd guess so since model-switching is pretty easy on ChatGPT these days

[analog magnet]

Is anyone else experiencing problems with ChatGPT? Here's what's happening on my end:

On my desktop (Windows 10), both the ChatGPT app and browsers have stopped functioning correctly. I can't use plugins or actions with custom GPTs, and even the standard ChatGPT and custom GPTs can no longer read PDF files and see images. They claim there's no file to read or that they can't access it.

Additionally, my custom GPTs are not accessing the knowledge bases I've provided, nor are they performing web searches as they used to. It also seems like the context window has been reduced significantly.

These issues occur both on the desktop app and in browsers. However, I’m not facing the same problems with the mobile apps. On my Android devices (both tablet and phone), everything is working perfectly fine, including plugins, actions, PDF reading, and web searches.

I’m a Plus subscriber, so I found this discrepancy between the desktop and mobile experience very odd. On desktop, it feels like a completely different (and less capable) model is running.

Is anyone else experiencing similar issues? Or does anyone know what might be causing this?

Over the past 4–5 hours, I’ve been dealing with persistent issues on my desktop (Windows 10). I’ve already reported a couple of these bugs, but the more I tried to use the desktop version, the more problems I uncovered.

While I’ve managed to submit two bug reports so far, I’m left wondering: Do they actually review these bug reports? I’ve sent them hours ago (about 4 hours), but there’s been no feedback or acknowledgment yet.

It’s becoming extremely frustrating because I prefer working on my desktop for certain tasks, and the current situation forces me to constantly switch between my desktop and the Android app, where everything seems to work fine.

[sturdy jewel]

It’s the afternoon before a holiday in the US, I don’t think you can expect any response until next week, would be my guess

[analog magnet]

I just want them to fix this, is everything normal in your ChatGPT ?

[sturdy jewel]

I haven’t leveled up to plug-ins and actions 🙏

[analog magnet]

but can your Browser and Desktop App ChatGPT read PDFs and Browse ?

[sturdy jewel]

Seems like it works

[strong charm]

Hi there, I'm a paying customer and have been experiencing mid-answer stalling from chatgps (be it the macos client or the web page). What gives?

[maiden folio]

on the Windows App, Its difficult to start a new chat with 4o or the other default models. I have to go into chat history and find one, and hit new chat. Because the left side panel, doesn't list it as an option, just custom GPTs.

[timber shadow]

If you close the sidebar and then click the new chat button, it should start a new default chat instead of a custom GPT chat. I bet this will be fixed soon!

[sturdy jewel]

If you’re using a VPN, test it with it off, some other people were having some problems with that

[dense hill]

why is chatgpt unable to describe images now?

[rustic stump]

Guys I can't read my professor's handwriting. Is there an AI or something that will read and make it a readable text?

I tried image to text generator but they don't work properly

[onyx kraken]

You can try to put an image of it uploaded to ChatGPT and see what it can make of the writing.

You can also talk to the professor and write down yourself what the teacher says, or with their permission record the speaking.

[dense hill]

Does anyone know why ChatGPT can’t see images anymore?

[rotund holly]

Hi everyone I have a question, so long story short I was having trouble with my OpenAi account and had contacted the support team, as you see they got back to me but I don’t exactly understand what is a “hard value”? 😭😭😭 sorry if the question seems stupid I am still pretty new to all of this, is hard value just like just setting a limit in my usage?

[sturdy jewel]

Yes, you got it, it’s setting a limit

[royal scroll]

Yes an expenditure limitation

Max limit 1000 usd, warning limit 500

That’s mine

[blazing crypt]

So i tried to add billing to the api but the payment got declined, i refresh the page and the payment was successfully added and I am at tier 1 with the $18 credits and 120 limit but because i wasnt sure if the payment was actually valid since it says decline at first i took it off to resubmit it and now it declines and doesnt add anymore.
support told me it could happen if it was prepaid which I do use that but then why did it went through/upgrade me to tier 1 initially? I still keep the benefits of tier 1 without being charged yet?

openai help chat has not responded to me since our initial discussion

[rotund holly]

Thank you!!

[strong charm]

nope no VPN 😦

[sharp fog]

Finding it really annoying that there’s no easy way to directly open a voice web search on the iOS app. As far as I can see, I have to use the web search shortcut, then click a couple of times to get to voice (non-advanced). Going direct (to advanced voice in 4o) doesn’t allow web access. Am I missing something?

[pure grotto]

I am new to AI, I want to build a application which reframes videos from 16:9 to 9:16 using dynamic cropping, subject detection. How do I do this? Any suggestions?

Ping me while replying

[finite blade]

Good morning, I have had a problem with GPT for a few days, in which it has not read/interpreted the images that I have been sending it and it has been like this for a couple of days, I would like to know if the same thing has happened to anyone. or what it could be due to.

I have chatgpt-4o

[slate dawn]

same thing i don't know if you did find a solution or not

[finite blade]

so far I haven't found a solution

[hard musk]

what is going on with chatgpt this week ? why does it say it can't do this now?

[hard musk]

I even have the pro ( been happening all week

[finite herald]

I am uploading images, but I am facing an issue where the message 'I still can't view the images' shows up when trying to create captions. Kindly explain this issue in more detail and resolve it.

[polar kernel]

Hey! I need help. In the view, when I want to edit my own GPT, the "create " tab has suddenly disappeared. Only „configure“ is possible in my view. How do I get the the create Chat back?

[marble lion]

I got the same problem. someone got a solution?

[kind ferry]

A small indie company OpenAI. At the same time, it works perfectly on my girlfriend's account without a subscription

[oak cloud]

My ChatGPT browser version is experiencing issues - both search functionality and image uploads are failing. However, everything works fine in the macOS application.
Whyy.

[fading bough]

quick question, why is chatgpt redirecting me to it's login page when it was working fine normally? I don't feel like logging in to ask what's the best option to get

[rigid glacier]

Hi, i have the premium plan but suddenly it started to tell me that ChatGPT cant read the files i upload or the photos. The size of the file is like 255 KB, so is not a heavy file. Anyone knows how i can fix that?

[hushed crane]

What is the problem here?

It has created 20 other characters thus far. For 4 of them it kept putting Star Trek and Star Wars IP in the pictures against my wishes and I had to keep telling it to redo those ones without that stuff. Now this one it says there's a content violation???

[onyx kraken]

Does it work if you remove the character name from the prompt?

[hushed crane]

Nope.

[onyx kraken]

Hrm. And replacing federation and battlemech perhaps?

[hushed crane]

It had no problems with any of the other characters using those words.

I had to trim it down to
"A woman in her eighties wearing a military flight suit."
"Redo in square aspect ratio please"
"Redo without american iconography"
And it still blended a granny shirt with the operator suit.

[onyx kraken]

I don't know which words are the specific problem.

However, this worked:

"A dignified elderly woman with white hair tied into a bun, wearing a futuristic, cleanly designed flight suit. The suit features subtle patterns and designs that evoke a sense of unity and teamwork. She stands in a futuristic command center filled with glowing panels and holographic displays, exuding an air of wisdom and authority. The scene emphasizes her role as a leader in a high-tech environment, coordinating and supporting nearby allies with advanced technology. The background includes a view of stars visible through large windows, adding to the futuristic ambiance."

Can you modify that into what you need?

[hushed crane]

I'm grateful for your help in that. At this point, I'm happy to just move on to the next character, this one has been so frustrating. Even "Old Woman in Flight Suit" got flagged somehow. Does it think a flight suit is the same as a cat suit??

I just don't understand how it can inappropriately put walkers from Star Wars in the background, no problem, but if my words ask for something even distantly related to some IP, it freaks out.

[pure grotto]

Please help me

[limber dome]

Subject: Issues with uploading and interacting with PDF files and images in ChatGPT Plus

Hello,

I’m a ChatGPT Plus user and I’ve noticed a discrepancy in functionality between the Android app and the web version on my PC (using Chrome). Here’s the issue I’m experiencing:

Problem Description:
On the Android app, I can upload and read PDF files without any issues.
However, on the web version of ChatGPT (on my PC using Google Chrome), when I try to upload a PDF file, the upload option isn’t recognized and I can’t interact with the file. Also, when I try to send images, ChatGPT does not read or review them as it used to.
Steps I’ve already tried:
I’ve cleared the cache and cookies in my browser.
I’ve disabled extensions and tested in incognito mode.
I’ve updated Chrome to the latest version ([Version 131.0.6778.86 (Official Build) (64-bit)]).
I’ve checked OpenAI’s status page, but there don’t appear to be any ongoing issues.
I’ve tested in other browsers and on other devices, but the issue persists.
Additional Information:
My account is subscribed to ChatGPT Plus.
This issue only occurs in the web browser (Chrome). I have no issues in the Android app.
I’ve attached a screenshot showing the error message when I try to upload a PDF or an image (which I can’t display here, but it is included in the email).
I would really appreciate your help in resolving this discrepancy between platforms and ensuring that I can access all the features of my Plus account on any device.

Looking forward to your response, and thank you in advance for your attention.

Best regards,
Jesus

[finite herald]

I am working on the captions for the images, describing everything visible in the images. I have 4-5 users that I have purchased, and the same issue is happening with all of them. When I upload images, I get a message saying that I cannot view the images. If this issue doesn't get resolved soon, it will be of no use to me, as I cannot complete my work. Please investigate this issue so I can finish my tasks.

[finite herald]

This has been happening to me for the past 3-4 days. Please tell me how I can solve my issue. No one is helping me with my problem, not even the help center. They say I will receive an email, but no email ever comes, and there is no response on Discord either.

@trim copper

[sturdy jewel]

Have you tried clearing your browser cache, logging out and back in, trying a different browser, restarting the device, starting a new chat, rephrasing how you’re asking what you want, etc. Sorry if you’ve already tried that. It does sound frustrating.

If it’s still a problem after all that I would search in #1070006915414900886 for a similar issue and add your experience

[raw oyster]

Hey guys I would like to know if someone know how I can overpass the AI detectors pls ? Because I have a enormous text that I made with different AI but like it s getting detected, and the AI humanisers aren t working that much in my case

[sturdy jewel]

Have you tried rewriting it in your own words?

[raw oyster]

No because I do not really have the time cause I have many exams and it s like a report so It s really huge even if I know this could be the solution

[sturdy jewel]

How do you eat an elephant? One bite at a time

It probably won’t take as much time as you think, just do it in 15 minute segments

[raw oyster]

lmao love the exemple tho

[sturdy jewel]

Old proverb, it’s not mine 😇

Set a timer and see how much you can get done in 15 minutes

[finite herald]

@sturdy jewel I have already tried all these methods, but there has been no change. The result is the same.

[cosmic mesa]

hi there

I recently facing a problem with chatGPT 4o, as I upload an image but it can read or recognize it, while I upload the same image with the same account on my phone and it works, but it doesn't works on my laptop

what should I do?

[sturdy jewel]

People have been joining the server and asking the same question the past few days… Hopefully they get it sorted out soon

[median thistle]

is the 4o image generation ever going to come out? I don't even care if the quality is worse than dall-e, i just really want the ability to generate multiple images with the same character

[flat comet]

Hi everyone, I have a problem with ChatGPT, it doesn't recognize the files I upload, and I have the paid version, can anyone help me please?

[cosmic mesa]

and no solution provided yet?

[sturdy jewel]

I’ve been recommending people do a lot of standard stuff like clear their cache, delete the cookie, try a different browser, restart the device, etc. If it does seem to work for them, they aren’t reporting back lol

[cosmic mesa]

i'm using the windows app. However I used the browser and it showed the same issue, but it works well on my phone idk why

[candid lake]

I tried all of this, cache, cookie, incognito mode, another browser, another device, another VPN, log out, clear memory. In one way I find to evade this problem, i run another account without plus subscription in my case, as usual log in, and it working normally

[cosmic mesa]

so it works on the free but no the paid version?

[candid lake]

I'm not sure, because in another text chat in the discord, the person says that everything is paid for and he uses everything normally

Perhaps this is a coincidence, or perhaps an unsuccessful subscription date that "infected" the account

who signed up/renewed subscription when?

mine is 25 nov

[brave walrus]

I have a new theory and it's gonna sound weird, but have you tried toggling your memory off? And... And trying again, or also... toggling your custom instructions off, possibly. I only mention it because I'm scrolling through X and people are seeing odd behavior when their memory is overfilled. And this would also kind of make sense to me because you have to try to shove everything into a certain context window. And if the model set context is just astronomically huge, then it's going to have like no way to fit that all in there... idk just a shot in the dark.

[onyx kraken]

Just an observation.

I haven't been affected by the bug. I also haven't let my memory reach 100% (or more if that's possible) in months.

But those who are at 100% memory it tells you (or did months ago) at the top of the screen.

It could be affected by memory or custom instructions for those affected, no idea. But memory full should be visible top of screen

[viscid mountain]

Hello, have you experienced GPT 4O chat and can't read images?

[wary totem]

i dont have the option to talk to my custom gpts, does anybody have this issue? i only have access to configure and preview

[brave walrus]

top right corner

[wary totem]

i have that same screen but what happened to the chat for custom GPTs

[brave walrus]

You can go to Explore GPTs, then My GPTs and find your gpt in there to chat... or just https://chatgpt.com/gpts/mine

[tribal laurel]

chatgpt is slow on chrome with long chat history but runs fine on phone ive gotten page unresponsive at times on chrome

like this lol

[candid lake]

hello everyone, today, as usual, I went into the chat to try to generate an image, and since my site did not open on my phone in chrome, I went incognito to check that it was cookies. So I enter the request, and chatgpt surprisingly fulfilled it. I was already glad that the developers had fixed everything, but the devil pulled me to check how o1 preview thinks. I open a new chat, select this model, and everything abruptly comes back, this whole problem. I tried to delete all cookies in this incognito mode, created a new incognito tab, even opened another browser in incognito mode, but everything remained the same

[marble lion]

I have this problem for a few days now and tried everything they told me to do. Now the team doesn’t respond to me

[tender bough]

None of these examples work: https://openai.com/index/hello-gpt-4o/

How is it possible that none of those examples does not work as expected?
I'm using ChatGPT Plus

I'm missing building window. Bug or how to fix?

[tender bough]

I have the same problem.

[tender bough]

Before we had ability to build GPT with conversation. Now there is only Preview-mode and conversation possibility is missing

[cosmic mesa]

have anyone tried to delete ChatGPT's memory and instructions to solve this -image and document recognition- and it helped?

or even deleting the previous chats

[candid lake]

i tried, deleted all memory and disabled instructions and memory usage

didn't help

[short lion]

Hello
i am having this error eveyrtime i try to use a picture but if i put the same picture in another account with gpt 4 it will work why?

[fast bramble]

Hm? I am new to this so I don't know many terms but if you are asking...to generate images on the free model, then yes, I created an image for my webnovel book's cover...though the qual6was not good, it was generated successfully...and more than 2-3 images

[wary totem]

i think its a huge bug, OpenAI doesn't even know about this yet

[marble lion]

I sent a message about it and the told me to try another browser, device and clear cache. I told them it didn’t work and they do not reply back to me for 2 days

[cursive patrol]

I am having issues with the Custom GPTs in ChatGPT.com - It looks like others are having the same issue (ICANY. and @wary totem). There used to be two buttons at the top of the page, a CREATE button and a CONFIGURE button. There is NO "create" button anymore. Is that by design or is this an issue.

[median thistle]

The current image generation uses Dall-E 3, which is a separate model that GPT 4o prompts to create an image when you ask for one. The generation I'm talking about is built into 4o, allowing it to take into context previous images when you ask for a new one. This allows you to generate multiple scenes with the same character and objects, combine aspects of multiple images together, ask for specific edits to images, etc. This was announced six months ago but they still haven't released it or even talked about it.

[short lion]

@rustic marlin @exotic fiber

[rustic marlin]

It seems to be a bug, some people reported similar problems

Try clearing browser cache

Other than that, just wait for oai to fix

[short lion]

Okay thank you
Tried that and it didn't work hopefully a quick fix because i have finals next week

[sturdy jewel]

They’ll probably look into it by Monday

[stable canopy]

Hey, im getting problems with sharing link with my gpt, and now somehow not able to read its content (google document) i have chatgpt plus, but i dont have access to plugins :/ wierd...

but like 1-2 days ago, this fuction worked... idk what happend

[brave walrus]

Oh, I know what you're talking about now. I remember that. You just described a GPT and it would fill in the descriptions and stuff for you. I remember that I didn't really find it useful though. I Wonder if they took it out because nobody was really using it or it was becoming frustrating. I don't know.

[stable canopy]

or document, or links... these was working like 1-2 days ago... now chatGPT can even read attached document... like what the hell?

its thanksgiving in the US and the chatGPT dying XD i dont understand

[dry crest]

Is anyone else Presently having issues uploading pdfs into teams account? Is there any way to get some help with this? This issue is happening with standalone app for windows, and chatgpt website on ms edge browser.

[viscid mountain]

sureee today i still same issues

[karmic whale]

Mobile app seems to be the work around right now.

[wary totem]

whats the point of creating GPTs if we cant get the GPT builder to work

[hasty bloom]

Hi all, anyone else experiencing issues with ChatGPT right now? Since 10 hrs or so I get answers, that are not related to my questions at all but seem to be from an old chat, my questions are not identified as questions at all and I receive answers like "Oh, I understand, so you are asking..." and when I ask "what is 2+2" in a fresh chat, I am getting "Error in message stream" over an over again. No matter if I use Mozilla on Linux, a clean sandboxed installation Mozilla, app for iPhone, or Mozilla / Edge on Windows.

[brave walrus]

https://help.openai.com/en/ speech bubble, lower right, select through to support there.

[plucky silo]

Hello i have been facing an issue with chatgpt since 2 days
it doesnt generate pictures or read documents or analyze pictures is there any solution to this and i have been trying to contact the support for the last 2 days but i cant reach them

[trail cairn]

https://discord.com/channels/974519864045756446/1312045865107460176

[plucky silo]

where is the solution i cant find it

[trail cairn]

there are none but that's the thread you should vote, post and follow

[plucky silo]

alright

[hasty bloom]

I did, however no one is responding. As I have not been asked to identify myself in any way, how do they respond when I shut down my computer? Stupid question maybe, however I never had to contact open AI support before...

[brave walrus]

If you logged in, your email. It should just be tied to your account.

[hasty bloom]

Thx... as they address me with my first name, I am seemingly logged it. Lets hope for the best... ChatGPT seems do degrade by the minute, it now has started to change languages mid-sentence, when telling me that my question didn't come through.

[brave walrus]

try toggling off your memory and custom instructions if theyre on... seeing if it starts making sense again in a new chat. if so, I'd prune the memory down.

[jovial summit]

same

[jovial summit]

same for me

[golden dove]

gpt-o1 not working!!!

Nothing doing!!! Clean cookie also not working

[brave walrus]

same what? there's a create link there. it works.

[hasty bloom]

was some of the first things I have done. freeing up memory as far as I could, also clearing cookies / cache. didn't change a thing. no custom instructions in place. I am just using vanilla GPT-4o. Also does not seem to be related to my local settings, tried different OS on different machines as well as on my iphone from wireless lan as well as from mobile internet.

[jovial summit]

cant modify on interact with my gpt builder

[brave walrus]

memory as in:

[brave walrus]

there is no builder. its just a config page where you define your GPT. dunno when the builder left, but its gone. you can select the image + sign to make an image with dall-e, etc. all functionality is still there. if you need help making instructions, just do it in chatgpt itself and copy paste? idk.

they probably need to update this page: https://help.openai.com/en/collections/8475422-building-gpts

Passed this along.

[hasty bloom]

did exactly this

[brave walrus]

All I can do is wish you luck with support. Sorry.

[hasty bloom]

Thx mate... right now I hope this is just some hickup and will solve itself... dunno when somone in support might have a look at it and if they will be able to find a solution.

[candid hull]

Using completions api, if you specify a json schema response format, should you also include the same schema in the system prompt? Or can it be removed to save tokens

e.g. system prompt:

api response format

[swift forge]

What's wrong with ChatGpt Premium? He stopped recognizing photos, files, and drawing pictures.

[trail cairn]

broken https://discord.com/channels/974519864045756446/1312045865107460176

[swift forge]

thx

[woeful minnow]

Why do the o1-preview and o1-mini models react without step-by-step thinking, and respond instead as the 4o-mini model?

[golden dove]

open ai dont want fix it

[fading robin]

o1 mini works normally for me

[golden dove]

not working :/

@fading robin where you frome?

wich brouser do you use

[charred parcel]

o1 mini is working fine for me using the macOS Desktop App

[fading robin]

EU. And Chrome. On PC. Windows.

[golden dove]

ehh...good.. Me not working(

[onyx kraken]

Happened to me last night for the first time.

I noticed a lot of problems:

selected o1 models both didn't show thinking and did have partial knowledge of what's in my memory, could tell me about some of it.

4o-with-canvas could not open canvases and didn't know about the tool.

4o itself couldn't see my full memory.

I was able to:

Use o1 normally in a different browser than the one I was using when the problems appeared.

So I:

Tried to clear cache (by itself didn't fix, maybe not needed).

Log out and in account (did not fix, maybe not needed).

Turn off memories and custom instruction, log out and back in again. This did fix. I have no idea yet if cache needed to be cleared too.

Recommend trying those steps, let us know if it's a fix for you too.

[golden dove]

not working

[echo temple]

hello - i'm having issues logging into my account on platform.openai.com. It's telling me i need to verify my email, but no such email ever arrives. I've tried emailing [e-mail], but I'm not getting a response. anyone have insight to this? i'm using sign in with apple, and I think that may have borked something

[charred parcel]

if you're using Apple Sign In then email password recovery will not work because the password is stored by Apple not OpenAI. To get further help you would need to contact https://help.openai.com

[echo temple]

guess i'll just submit a bunch of tickets until someone replies lol

[tender brook]

hello, I have x2 chatGPT accounts under my credit card for me and my son. My son made the mistake to use his former company email as his account with no means to log in again to recover his password. how can I either 1) have his email account changed to update to a new more permanent email. or 2) close/cancel my son's account and create a new one without me 3) going nuclear and cancelling my credit card, letting my son's account die along with an old credit card while I refresh my account with a new one. Is there a life person phone number in ChatGPT customer support who can help me with this? the help chatbots are... limited

[vagrant plover]

There is no customer support helpline. You should get in touch with https://help.openai.com/ for additional assistance, you should be forwarded to support for billing issues.

[inner pawn]

Every time I open Chat GPT it crashes Chrome immediately. (Chrome version 131.0.6778.86 on a 2019 Mac running Sonoma)

Doesn't matter if I clear cookies, etc. still crashing Chrome. It's not crashing Safari.

https://help.openai.com also crashes it.

[tender brook]

thanks. looks like the help.openai.com chatbot has some escalation. I got a "the team will get back to me in 3 days. there is hope 🙂

[cold folio]

I want bulk delete for gpts. and a better search feature. and probably a recover from deleted system that waits 30 days otherwise the gpt will be forever gone.

also order by most used to least used gpt

then i just delete the gpt i never used. yeah

[cold folio]

I wonder if OpenAI ever thought on switching from Intercom to some other support platform

[proven charm]

Hello all, paid subscriber here.

I’m not usually one to talk inside one of these offical community things, so this is mostly a first for me.

So, for the past few days my chat history for some of my most recent sessions on Chatgpt have been disappearing as of late.

All of the times it has happened is during particularly lengthy sessions. At a certain point after I’m done for the day with my gpt, I return to my prompts being missing.

Not only does it delete many of my prompts but all of the history of it’s responses, giving me the ‘There was an error generating a response’.

It has been incredibly tiring and irritating, as I’ve been using it to help create stories.

I don’t know if this is a bug or not. I’ve cleared my browser’s cache but the old prompts and responses remain missing.

I have no clue what to do about it, and I’m wondering if anyone else has happened to experience this

[onyx kraken]

I'm curious.

When you say it's deleting a lot of your prompts and all the history of the responses - do you mean that you're coming back to 'the chat you had' - only most of it is missing, showing some of your inputs and none of the model's outputs? (I've never heard of this happening to anyone)

Or do you mean a bunch of entire chats look like they are missing? (this can happen if someone gets into your account and deletes some chats, change password and contact https://help.openai.com/ to discuss the problem, eek)

Or do you mean it's like the chat 'went back in time' and 'stopped early', but everything up to that point is there? (that can happen if an earlier input you make is edited. This creates branches, you can click the arrows to move between them. But while on the new branch, the old branch(es) are hidden until you use the arrow, it can look like the conversation after that point is gone.)

Or something else?

[proven charm]

"Do you mean that you're coming back to 'the chat you had' - only most of it is missing, showing some of your inputs and none of the model's outputs?"

Yup, exactly what's been happening. I've never had this happen before until these past few days. I've had many long sessions in the past and none have ever started deleting prompts and responses at a certain point. Memory loss? Eh every now and then, but never this

[onyx kraken]

Sounds weird and fascinating. I'm sorry that's happening to you, is there an affected chat you'd be willing to share so the rest of us can see?

The 'good' news is I have no idea how anyone could cause this to happen, doesn't sound like someone's getting on your account and doing stuff.

https://help.openai.com/ and bot bottom right, log in if needed, click through the choices until you can 'report a bug'.

I'd definitely share with tech support shared conversation(s) that are affected, so they can see the problem and see which chats to consider to try and understand what's going.

Sharing a chat here is just so the amazed and curious can see something we've never seen before (least I never heard of this happening to anyone).

[proven charm]

There isn't a really good example that I can use sadly. All I do know is that at a certain point in length, it happens. In my most recent chat with a gpt I created, I'd say I got to around 150-200 prompts in total (which is mostly from me updating story prompts to get the best continuation for my story) before it happened

[proven charm]

Ahhh okay, I'll definitely try to look more into it then. I appreciate it!

Oh and believe it or not, people have had similar problems, a bit similar to my own at least from one of the online forums I read through after a brief search earlier.

A few links if you want to read through them:

1. https://community.openai.com/t/chatgpt-responses-disappear-the-next-day-and-must-be-regenerated-web-app/662525

2. https://community.openai.com/t/chatgpt-losing-messages-in-middle-of-conversation/676852

[onyx kraken]

Interesting, thanks. First I saw anyone discuss them on the discord.

[proven charm]

Of course, happy to have been able to give more insight

[quasi willow]

the chat history export hasnt been working for a few weeks now. anyone else with the same kind of problem?

[fringe heath]

openai support not respond at all

i've sent them message since 26 hours ago

[brave walrus]

That looks like a whole lot of support to me actually. They've indicated that their system shows you on a free plan. Doesn't matter what you screenshot them, if you don't get that fixed, nothing's really going to work right. I would definitely try working with them instead of trying to meme on them. These people are your only lifeline. They're not there to argue with you. They're there to try to get the information out of you that they need to help you. They also have a lot of people that they need to help. So it might take a while for them to respond. Please be patient.

[fringe heath]

it's been 3 days for me

[brave walrus]

That's not uncommon, sadly.

[fringe heath]

many people reported about this , when they planning to fix this

[sturdy jewel]

It was a holiday weekend in the US

Everybody still sleepy from eating turkey, but they will be back at work tomorrow

[brave walrus]

I'm not exactly sure what issue you're referring to without going back through the conversation, but there are a few bugs at play. Some people have managed to resolve similar problems by trying these steps:

Clear browser cache.
Log out, then back in.
Toggle off or clear custom instructions and memory.
Repeat clearing cache, log in again.

Basically, it’s the typical troubleshooting loop until it works. If you're dealing with an image-related problem, similar steps could help. Images are just base64-encoded strings, and adding them to an overloaded context can easily cause issues, especially if you’ve got lots of personalized settings eating up space. If your system instructions, model set context, and custom instructions are maxing out your context window, adding an image could push it over the limit. Trimming down custom instructions or memory might help prevent overflow—just a theory, but it’s worth a shot.

[lunar egret]

Hi!, I'm currently facing this issue in several situations, I was wondering if any of you had the same error and howd you solve it. Im a plus user btw.

[sinful granite]

i have the same problem

[lunar egret]

Seems like using it in phone or mobile app works great

[brave walrus]

https://help.openai.com/en/ <- in the lower right there is a bubble. Click through to talk to support. I'm going to guess that openAI is a very reactionary company and will do little to nothing on issues they don't perceive themselves or see as a large problem via reports. If people are simply waiting for a fix instead of reporting something that is not shown as an issue on the status page, then there's not much for openAI to react to.

[fringe heath]

they won't even reply 😂

i've messaged em and got no reply 28 hours ago

[brave walrus]

I've gone through support before to get problems fixed. I know it takes a long time. You have to be patient and persistent.

[woven rock]

So uh- ChatGPT is straight up refusing to search the web, like- At all, it keeps saying it can't, when I know that it can

[charred parcel]

it's working fine for me in the ChatGPT macOS Desktop App

[woven rock]

Gosh darn GPT gaslighting myself and itself

[primal mantle]

So, looks like it wasn't just me, All models are working as the free model, I'm trying to contact support, have you guys made it?

[charred parcel]

to contact support use the chat at https://help.openai.com

[primal mantle]

Aight but that doesn't help much xd, bc that says "use your account to contact support" aaand

as you can see, theres no "Help" button

or any way to contact support

[charred parcel]

the only way to contact support is use the chat at https://help.openai.com/

[primal mantle]

I mean...

It is a dead-end circle

[charred parcel]

you don't contact support through ChatGPT you go to use the chat at https://help.openai.com/

[primal mantle]

step 1

[primal mantle]

step 2

[charred parcel]

if you go to https://help.openai.com/en/ in your browser then on the bottom right is a chat bubble, use that to contact support

[primal mantle]

should they answer back, right......... right?

[charred parcel]

sometimes in a few hours sometimes in some days, just lodge your support request and then wait for a response

[sinful granite]

true but i usually work on my pc so i find my subscription useless 😦

[lunar egret]

Me too, but to make it useful and avoid loosing my money

I send the images to my phone from one way or another and search it through there

[sinful granite]

will they fix it in the upcoming days?

[lunar egret]

I hope so

The unique way to let them know this is a problem thats affecting part of the community is reporting the error through help page

[candid lake]

My account finally got fixed

And on mobile browser it works now normally

But after logging in from the computer, everything remained the same and after that everything in the phone also broke down

Brilliant

[woeful minnow]

not working

[wintry echo]

#1313063703746318358 message

Ugh, when did they remove all the intended line breaks (eg: code ) from pasting into the input? This destroys workflow for devs

Feels strange knowing this probably went through multiple QA devs and reviews without a single person thinking, "Oh... this may be bad"

[charred parcel]

OpenAI has had reports of a few problems over recent days. If it's a bug then #1070006915414900886 is the right place to report it

[wintry echo]

Hopefully a bug :/ thanks!! I'd feel better knowing it's a bug and not a feature to save space on pastes 😛

[sudden spindle]

HEY
umm i need some info to enter here

from this i can add chatgpt to my sidebar

so pls help and give the info reqd in the screenshot

[charred parcel]

this is not an OpenAI product, please use #ai-discussions

[scarlet owl]

How can I get my prompts to give consistent results?

[charred parcel]

provide an example?

[scarlet owl]

I recently made a prompt to generate me notes in Obsidian in a aesthetics style for easier readability and access

My problem is, the AI doesn't keep a consistent style when I introduce more content to it, to make notes from

[heavy patio]

I suggest asking the nerds over in #prompt-engineering. They are generally pretty friendly and helpful, especially when you are willing to share the whole prompt

They're some of my favorite nerds, although the nerds in dev's corner are generally top tier also

[scarlet owl]

do you happen to be one of those nerds by chance

[heavy patio]

Sometimes I try to hang out and pretend like I belong with those geniuses I don't think they've sensed the imposter, yet . (Or they are just very kind)

[vestal dragon]

ChatGPT-4o is not recognizing images

or .pdf, .docx files

[charred parcel]

are you on a Mac or Windows?

[quasi willow]

did you manage to get the data export working? still isnt working for me

[charred parcel]

click on Configure in the top left hand pane

[sturdy jewel]

ChatGPT can make mistakes, the TOS talks about this, for example

When you use our Services you understand and agree:

Output may not always be accurate.

https://openai.com/policies/row-terms-of-use/

[charred parcel]

if you want a refund then read this -> https://help.openai.com/en/articles/7232895-how-do-i-request-a-refund-for-chatgpt-plus

[vestal dragon]

windows

[shadow plume]

I have GPT Plus, but for the past 3–4 days, I've been unable to use any features on my computers. No matter which model I use, I always get responses from GPT-4 (October 2023). It works fine on the mobile app, but I can't use it on my computers.

I’ve already tried all the basic solutions, such as:

Updating everything
Removing ad blockers, VPNs, and browser extensions
What else can I do to fix this issue?

[shadow plume]

i think this guy had my same problem....

[shadow plume]

same problem as this guy

[shadow plume]

same problem also

[shadow plume]

same... i think is a common problem...

[woven rock]

Heyo, Anyone suggest a good Custom Frontend I can use with the OpenAI API?

[vale forum]

With what goal?

[woven rock]

Mainly for using o1-preview

It's been such a help and I've run outta uses on ChatGPT

[vale forum]

No, since o1's response is kinda unique, there arent' a lot of off the shelf UI's that cares to such a thing

[woven rock]

Ah-

[vale forum]

You could try OpenWebUI and see if you like it

But it probably won't be as smooth as chatGPT

[woven rock]

Gotcha

[keen valley]

Hello, if I were to interigate chatGPT into an website, what would the cheapest and give me the most flexibility way be?

I plan to create applications or services using AI with custom instructions made by me earlier

[vale forum]

you'd create a service using API's, you wouldn't using "chatGPT" itself

the rest of the question is too vague. Define cheap, most flexibility etc

cheapest way is to host local model and serve it from your pc. that might not be the most desirable

[keen valley]

The plan was to use chatGPT to answer their questions and chatGPT answer would then be displayed on my website

would that be utilizing API?

[vale forum]

Yes

you can embed "chatGPT" the website into your website.

You'd use the API to create your own UI on your website

[final olive]

I believe that should be "can't" embed chatGPT.

[primal mantle]

Was it automatic or did you have to contact support?

[weak shell]

I can't login through the app, but the website works (although slow)

[tawny adder]

I purchased the ChatGPT Plus plan for $20, but the features aren't working (it's not generating images, not reading files, and the responses are outdated). I've tried all models, can anyone help me with this?

[candid lake]

Automatic

[crimson canyon]

Yep, same problem here. No export emails for at least the past week, with multiple attempts per day. This despite an indication from both the website and the app that the export was successful and I should receive the email soon.

[dire cave]

I don't know if I'm in the right place:
i work with the openai dotnet package and would like to use “tools”. I have OpenAPI code for this. How can I integrate this into my C# code so that it is also used by my code?

[charred parcel]

ask in #dev-chat

[dire cave]

thx

[cerulean mesa]

the features are broken for several days already, the only workaround are using free account or mobile app. No fixes or anything whatsoever rn -,-

[eternal oracle]

And no news from OpenAI ._.

[ionic frigate]

Hello, so if im not wrong, the chatgpt plus plan is not working well right?

[limpid flare]

yeah, and theres no answer from openai

[pastel mulch]

Hey, ChatGPT stops working for me every now and then (just doesn't load at all) any ideas?

[timber shadow]

Hey! Have you already tried clearing your browser cache?

[pastel mulch]

Yea, it works now, seems like this only happens in my edge profile, ty for the help

[modest creek]

ı have same problem too

[eternal oracle]

same

[trail cairn]

star and post https://discord.com/channels/974519864045756446/1312038815321428021

[wary adder]

why is chat gpt not reading images anymore?

[fringe heath]

https://discord.com/channels/974519864045756446/1312038815321428021

[fringe heath]

and still not resolved lol

[buoyant seal]

Yes

[celest pike]

How can I get ChatGPT to achieve exact word counts (e.g., one topic with 200 words split into 2 x 50 words and 4 x 25 words) using Python for precise counting, with correction passes until compliance? Despite including Python in the prompt, consistent results have been elusive after days of attempts.

[charred parcel]

you can't, the models are language models not maths models and are hopeless at counting

[celest pike]

thanks for your answer. But when i ask chatGPT after the job is done to count with python i get 90% the correct result. Only during the creation its too unprecise. Itteration often doese not realy help. No other ideas? Or maybe a hint to an othe ai?

[charred parcel]

all models regardless of who makes them will suffer the same limitations

[celest pike]

so no need to doublecheck with api? It will be the same? 😦

[charred parcel]

the API will do the same

[celest pike]

thanks for your quick response, have a nice day

[wanton temple]

Hey a couple of days ago i couldnt acces gpt through my browser. Downloading the app couldnt fix it since im on a mac with an intel core. I emptied the cache of the browser, updated and switched to another without succes. When i try to prompt there its just pending. I mainly use it for shortening text and brainstorming in scriptwriting and have found it helpful. Would love to get it up and running again. Any suggestions on how?

[charred parcel]

are you using Safari?

[wanton temple]

At first firefox, then safari. The weird thing is that i have different issues. In firefox the chat history will not show and i get not response to my prompts. In safari however the chat history is present, i get an answer from my first prompt, but im unable to scroll down into the conversations. Been using it actively for 2 months now and this is all new.

[charred parcel]

what version of Safari are you using?

[wanton temple]

Version 15.6.1 (17613.3.9.1.16)

[charred parcel]

on which version of macOS?

[wanton temple]

12.6.5 Monterey

[charred parcel]

Is there any reason you haven't updated to Big Sur?

[wanton temple]

isnt big sur the previous?

do u mean ventura?

[charred parcel]

yes sorry, I got confused

[wanton temple]

No reason, a quick research made me think i couldnt on a mac book pro from 2016

[charred parcel]

hmm, your version of Safari is very old now, have you tried Chrome?

[wanton temple]

ill try it out hold up!

[charred parcel]

no worries, I will be here

[wanton temple]

switching to chrome solved it! thx!

[charred parcel]

great, have fun 🙂

[somber echo]

please help chatgpt does not process files and images, does not give errors, writes that it cannot process inventions in the response message, Plus subscription is paid, please help

[ionic frigate]

its a general issue. we are all in the same place, gpt is not working as it should

[somber echo]

Okey, thank you 🙂

[slender island]

logout,and login again

[vale forum]

Sounds like ak nown bug, submit a bug report

[lusty lodge]

maybe you can stop the subscription then logout, and try to login and start the subscription again

Its worked for me

[tawny adder]

It's still the same, I've logged out and logged into the account several times.

[candid lake]

I understood what the problem is, right now, if I use a vpn in the USA, the chat cannot generate images and analyze them, and o1 does not think, but with a vpn in France, everything works for me so far

[candid lake]

in some regions, it will most likely work, and in some it will not

[fringe heath]

Guys it's been a week ago ...

Openai what are u doing

[fervent void]

wait what? i can generate images with o1 with french ip?

i just always assumed it cant do it

[fringe heath]

Really?

On windows?

[fervent void]

that's what @candid lake said above no?

[candid lake]

Not o1, but 4o generates images. And o1 thinks normally

[fervent void]

oh man i already had my hopes up to analyze images with o1

[candid lake]

Patience, my friend, it depends on openai, I think soon a full-fledged version of o1 will be able to do everything that 4o does, and even better

[fringe heath]

Dude its been 1 week ago what patience u saying 😂

[candid lake]

I'm not talking about the problem with file recognition and generation and everything else, o1 basically does not know how to create an image, the 4o model does this

[fringe heath]

I don't think there's will be much different about image genertaion cuz they gonna use same dall-e

Dunno let's see

[vale forum]

i dont think 4o generates images

I think it gets routed to dall-e for image generation

[candid lake]

I generalized because it has access to DALL-E unlike o1

[olive furnace]

it needs the ability to see dall-e output & understand the limitations of the drawing engine

[simple acorn]

Hey guys please help

ChatGPT has been acting weird lately. Even though I have the paid plan

it's telling me stuff like "Currently I can't generate images with DALL-E"

"Currently I can't browse the web"

However at least my ChatGPT APP is working properly on my Iphone.. but on the browser it's not working properly.

}Does anybody have any news about this?????

[fringe heath]

https://discord.com/channels/974519864045756446/1312038815321428021

[rugged wing]

something went wrong please restart the app and try again chatgpt login problem
how to solve this

[sour plaza]

Hey! i'm using gpt4o to translate a text and sometime i get repeating characters at the end of the response that make no sense (something like on the image). from what i understand this is the effect of negative frequency penalty, is there anything i can do to prevent this from happening?

[vivid mortar]

Modifying the temperature can help with this as well as adjustments in the system prompt like "once you have completed translating the provided text provide the end token and say nothing else" etc

[fringe heath]

chatgpt is working now finally!!

[serene thunder]

not for me haha, 4o still isnt reading full prompts, still cant search, and is still not as smart

[viscid mountain]

is still chatgpt plus cannot read images?

[patent linden]

yes

[vagrant sky]

Hello guys, sorry to disturb you all; I need to create a basic chatbot using vanilla javascript. I don't really know how I should proceed, can someone explain please; what I know is that I need to use an api key but where do I get this api key please

Are there any documentation to read on the openAI website to build the chatbot? I just register on the website but I'm kinda lost :c

[charred parcel]

this is how to get started https://platform.openai.com/docs/quickstart

[vagrant sky]

thanks !

[rugged wing]

hello
How long has it been since our GPT says that I can't analyze my photos
What should I do?
​

I received the same message from the support and I checked everything and I still have the problem:

Hello,
​
Thank you for reaching out to OpenAI support. Our apologies for the delayed response, and we appreciate your patience.
We understand that ChatGPT is unable to analyze your photos, and your concern about resolving this issue is valid.
The error you've experienced could be due to a range of factors. It might be an issue on our end or possibly a problem with your internet connection.
Here are some steps you can take to try and resolve the issue:

1. Retry the request again later - this often solves the problem.
2. Visit our status page to check for any active outages that might be temporarily impacting the service.
3. Clear your browser's cache and cookies.
4. If you're using a VPN, try disabling it.
5. Try using an incognito browser or a different internet browser/computer. Sometimes, a security add-in or extension can cause this type of error.
6. Disable any cookie blockers you might have enabled.
7. Consider connecting to a different network, such as switching from WiFi to a cellular/mobile network
   Even without a reported outage or issues with your network, these errors can randomly occur. While they're less common for ChatGPT Plus subscribers, they can still occur from time to time. In most cases, retrying the request later resolves the problem.
   We hope this helps! Let us know if there's anything else we can assist you with.
   Best,
   KC
   OpenAI Support

[trail cairn]

add star and post https://discord.com/channels/974519864045756446/1312038815321428021

[sour plaza]

my temperature is set to 0, and i added what you said in the prompt but it still happens sometimes unfortunately

[simple geyser]

@trail sequoia

[trail sequoia]

should I just describe the problem?

[simple geyser]

FIX YOUR DATE AND TIME

That's why you are getting the error

@trail sequoia

It because your date and time is wrong

???

Are you there?

[trail sequoia]

wow!

how can I thank you

tysm!!

[simple geyser]

Nah

[trail sequoia]

highly appreciated!!

[simple geyser]

Welcome

You don't need to thank me

[trail sequoia]

thank you too @timber shadow

[simple geyser]

Cool

Now do you need help with anything else

Or can I go?

[trail sequoia]

not anything rn

[trail sequoia]

sure thing!

[simple geyser]

If you need help dm me

[trail sequoia]

alr

I sent you a friend req so that I dont forget your acc name

[simple geyser]

Ok

[trail sequoia]

thanks, again!

[vagrant sky]

Hello guys, I'm trying to set up my first api request using the openai api but I'm having an error of quota, I change the model to the gpt 3.5-turbo and still gets the error, does someone knows why please

[charred parcel]

you need to add credit to your platform.openai.com account

[vagrant sky]

hmm I guess it isn't free?

[charred parcel]

no it is paid

[vagrant sky]

Yep I see, thanks !

[pine thorn]

Why do I have access to o1 pro mode already?

[honest nebula]

chatgpt it ma

[short patio]

do you have the pro plan

[boreal oxide]

hoping a ChatGPT employee can flag this. The paperclip to attach files appears for o1 pro, which was why I bought it. It doesnt work. Similarly if you start with 4o and upload a file, it then allows you to switch to o1 models, it then errors.

[sharp ferry]

my pro plan is not active on my account

[icy maple]

Is image generation broken again, or have I somehow run afoul of some insane content moderation policy?

What does this actually mean?

It seems the image generation tool is currently unavailable. If you'd like, I can assist with refining the prompt further or provide additional support for your project. Let me know how you'd like to proceed!

[fringe heath]

yeah it's broken again -_-

dunno whats going on with openai

[vestal dragon]

In the PRO version, did they fix the bug where chatgpt didn't recognize that I sent an image or document to it?

[viscid mountain]

still same problem

[cerulean juniper]

is there any way to export data from team plan?

[fierce ermine]

Hello anyone develop custom actions for the custom GPT to grab information from google driver that has some of my knowledge base PDF? I am only limited to 20 file uploads but I have a lot of documents that I want the custom GPT to get info from. Is this possible? Like can I make a schema for it? any help would be appreciated.

[short lion]

Im sorry but thi sis unacceptable its been a week and a half since gpt cant analyze photos

[lean fiber]

Hi guys
Around a few days when I upload my screenshot or my PDF ChatGPT told me he cannot access and ready image or pdfs but last month I don’t have any issue. I use the Firefox browser. Any limitation? I also mentioned that I don’t meet any limitation per day as well.

[charred parcel]

you willl need to contact support at https://help.openai.com

[charred parcel]

you willl need to contact support at https://help.openai.com/

[lean fiber]

Through your link I go directly to the message. I just only left my email and my question so they will be reply and solved my problems?

[charred parcel]

yes they should reply but they will be very busy and may take some time to respond, even up to a couple of days

[lean fiber]

Og it so long, is there any other solution its bug or happens only for me or any other solution?

[charred parcel]

are you on Windows?

[lean fiber]

I use mac, Firefox browser

[charred parcel]

ok great, download the macOS Desktop App for ChatGPT and use that instead

[lean fiber]

Ok thanks.
I try another browser and it works, i think it might from cache or something like that. As i had some important page on Firefox i didn’t try clear cache to see result. Thanks anywat

[charred parcel]

ok that's good news

[neon shell]

Hello, how can I choose the model for a custom GPT or it isn't possible

I ask this because I would like to try to create a custom GPT (for my use only) but with the new 4o with canvas

[charred parcel]

you can only use GPT-4o (no canvas)

[flat horizon]

anyone else facing this problem where chatgpt cant read any photos or files?

[trail cairn]

add star and post https://discord.com/channels/974519864045756446/1312038815321428021

[flat horizon]

thanks, hope they do smth about it

[simple geyser]

Hi guys

@rustic marlin

How to be mod?

I have helped people

[rustic marlin]

from time to time, we recruit for staff members, when that happen we post an announcement with a form for interested people to fill some info
there is basically no requirements except being 18+ and having a history of good activity here

[scarlet bluff]

can anyone explain why Start now for Chatgpt Pro loads Upgrade to team?

is pro upgrade not available yet?

[timber shadow]

Hello, it looks like you're currently on a Team subscription - is that correct? If so, as a test could you try navigating directly to https://chatgpt.com/#pricing? Once there, do you see "Personal" and "Business" tabs? I think the reason it might not be showing in the screenshot you provided is that Pro is considered a "Personal" plan, so it is included on that tab.

[scarlet bluff]

swapping to personal from business lets me click on pro. is there still api access through pro on personal?

[timber shadow]

ChatGPT subscriptions are billed separately from API usage, I am pretty sure this applies to Plus, Team, and Pro but I could be wrong. But you can still access the API with your ChatGPT account at platform.openai.com, you'd just have to load credit separately to use it.

[scarlet bluff]

ok, I follow, thank you

[fringe heath]

my chatgpt is not working well on my windows only works with my phone -_-

[nova smelt]

I have the teams subscription, which ends on January 11th 2025. I cancelled the subscription, because I actually I do not need a team license, but I want the Pro license.
When I click on my account, I can only upgrade my team license, I can not buy the Pro license. What can I do?

[nova smelt]

I do not see tabs, just this

[timber shadow]

Did you merge your personal workspace into your Team workspace? Or can you still switch back to your personal workspace? I wonder if the option might be there.

If you merged, I'm not sure what your best option might be since I'm pretty sure the merge isn't reversible. help.openai.com chat might be able to help

[nova smelt]

I think this is merged

[nova smelt]

I do not find this

I am in the help center and can only open the AI help stuff, but no chat

[timber shadow]

Do you see the little circle in the bottom-right of the page?

A small chat widget

[nova smelt]

Yes, just forwards me to help.openai.com

But I do not find any option to start a chat, that's why I came to discord now

[timber shadow]

On help.openai.com, do you see this in the bottom-right?

[nova smelt]

fond it. Now it has a "Message" button. It did not have before. Weird

[timber shadow]

I just found this: https://help.openai.com/en/articles/9793128-what-is-chatgpt-pro

Can I update my ChatGPT Team account to ChatGPT Pro?
We do not currently support updating from Team to Pro.​
Users with ChatGPT Team accounts can cancel their account and update their account to ChatGPT Pro once their subscription ends.

[nova smelt]

once their subscription ends.
This is not nice ...

[orchid rapids]

What are the best techniques and tools to have the model 'self-correct?'

CONTEXT

I'm a noob building an app that analyses financial transactions to find out what was the max/min/avg balance every month/year. Because my users have accounts in multiple countries/languages that aren't covered by Plaid, I can't rely on Plaid -- I have to analyze account statement PDFs.

Extracting financial transactions like ||||||| 2021-04-28 | 452.10 | credit ||||||| almost works. The model will hallucinate most times and create some transactions that don't exist. It's always just one or two transactions where it fails.

I've now read about Prompt Chaining, and thought it might be a good idea to have the model check its own output. Perhaps say "given this list of transactions, can you check they're all present in this account statement" or even way more granular do it for every single transaction for getting it 100% right "is this one transaction present in this page of the account statement", transaction by transaction, and have it correct itself.

QUESTIONS:

1. is using the model to self-correct a good idea?

2. how could this be achieved?

3. should I use the regular api for chaining outputs, or langchain or something? I still don't understand the benefits of these tools

More context:

• I started trying this by using Docling to OCR the PDF, then feeding the markdown to the LLM (both in its entirety and in hierarchical chunks). It wasn't accurate, it wouldn't extract transactions alright
• I then moved on to Llama vision, which seems to be yielding much better results in terms of extracting transactions. but still makes some mistakes
• My next step before doing what I've described above is to improve my prompt and play around with temperature and top_p, etc, which I have not played with so far!

[eager peak]

Can I somewhere see how many o1 prompts I have left ?

Need to know that before I start to work

[regal lynx]

Do you have ChatGPT Plus or ChatGPT Pro?

[sturdy jewel]

The other day I was using it and a little pop-up said I had 25 left

[eager peak]

Plus

[eager peak]

Yeah I know but I need actual number I have left

[regal lynx]

I dont know of a way to find the exact number. I just know it's 50/week

[sturdy jewel]

Yeah, I think they let you go over if usage is down maybe… so we might be stuck with this

[vestal dragon]

Imagine paying $200 to not have access to image reading

[cold folio]

I think thats a bug. i got it here

Mine is 20 usd plan

[fringe heath]

this bug can be fixed or need to wait for openai ?

[icy cedar]

same issue here

[vestal dragon]

It's not working here :/

[cold folio]

Have you enabled the additional features on settings? code-interpreter etc

try saying "use your image vision feature"

[thorny crown]

How much time do plus users get a day of advanced voice mode?

[sturdy jewel]

it’s around an hour

[thorny crown]

@charred parcel is o1 pro significantly better than o1 for coding?

[charred parcel]

yes I believe for coding radio theory applications it is better, for simple apps just standard GPT-4o is not too bad, o1 is good but for very scientific stuff definitely o1 Pro

[vestal dragon]

Yes

[little cedar]

is there a support channel? I am having trouble upgrading to pro

[sturdy jewel]

Unfortunately, you can’t get that sort of direct help here, but you can open an issue or contact Support through the app or website

[little cedar]

oh dang, okay thank you for responding fren

Is anyone else having issues upgrading?

oh i see there are people with this issue

and bugs have been submitted

[sturdy jewel]

I forgot, you can go here and click on the little chat https://help.openai.com/en/

[little cedar]

oh lovely thank you. I was able to submit a request for assistance. Hopefully they can help me out soon!

[stable canopy]

Guys... The image reading function is bugging again? I uploaded pictres 1h ago... now its not supported... What the hell... -_-

[north shadow]

what is going on? I get a error every time i try to ask it something. Why does it not work?

[compact sage]

What happened to uploading images? Am I the only one having this issue, or is it affecting others too?

[charred parcel]

https://status.openai.com <-- it's being worked on

[fringe heath]

well same thing ..

bruh

[charred parcel]

you just have to continue with support until you resolve it

[fringe heath]

They reply so late

[charred parcel]

you have no choice but to go through the process

[thorny crown]

would it be worth it to get if im using xcode on mac to develop ios apps and develop android apps?

[charred parcel]

possibly, depends on your needs

[dense obsidian]

Hello, I'm Gala I'm new here so I'l be certainly rude and unpolite and out of the good place, but I search someone who could help me about the platform Playground ? They decided roughly to abandon their "artisitic" branch to focus on AI design and graphism ... so I'm one tha t might make my suitcase and go away .. BUT they told me it was on my own to bring back my generated images, and it's unbelievable, because the only way they provide is to download 100 pictures at once ... I have thousands, for more than one year, I made series of illustrations, I had no place to stock them so they offered the "cloud", I used it. Now I have about one month after the Christmas familial festivities to pick up 100 by 100 images night and day .... I won't be able and I can't , I don't want to let any images lost, I have arts on way.
Does anyone know if there is one way to download my images faster ? Am I the only one in this case (such a donkey one )? Thanks in advance ...😢

[charred parcel]

can you succinctly explain what you need to do because I don't understand

[spark ivy]

yo
can u help me pls
chatgpt stopped viewing images
I tried all the models available to me
I have chatgpt plus

[trail cairn]

add star and post https://discord.com/channels/974519864045756446/1312038815321428021

[simple geyser]

hello guys

[onyx kraken]

What I understand:

@dense obsidian has some huge number of images generated through Playground and stored somehow in an OpenAI cloud service (or they think that's who owns where it is stored).

Gala seems to understand that this storage option is ending soon (perhaps because Gala is changing services, perhaps because the storage is no longer going to be offered). Gala wants to download all the images efficiently but sees only an option to download them 100-at-a-time, and perhaps in a way that isn't efficiently allowing the next 100 to be easily and rapidly selected.

Gala wants tips/tricks/methods to get those images downloaded easier and faster.

Gala, is that correct? Also, why do you think you lose access to the storage soon?

[trail cairn]

the weird part is that you cant create images at the playground and image links generated via api are valid only for one hour

[onyx kraken]

Yep. And labs only allows 1 download at a time, right? So I don't think Gala meant that either. Maybe some 3rd party option's involved.

[crystal ocean]

Chat make me video is normal ?

[vagrant plover]

It's able to edit and create videos through code execution in python, though it's not perfect - are you experiencing any specific issues?

[crystal ocean]

yes problem with generating in 3D or 6D , 8D . Camera is only 2D

[sudden tendon]

Hey brothers,
I’ve been having an issue where ChatGPT seems unable to properly analyze or interpret images I upload. It’s not processing them the way I expect and I’m getting inconsistent results.

Has anyone else experienced this? Any suggestions or fixes would be really helpful!

[reef night]

Hello, I use Google Mail to log in to chatgpt, I have purchased a chatgpt membership, and I have linked to openai DALL·E 3, but I cannot directly generate images in my chatgpt. I need to generate images directly in chatgpt. Please help me solve this problem. Thank you.

[cerulean mesa]

no fixes, it's been more than a week now

only mobile is working for me

[thorny crown]

why on the gpt pro plan it says i have to wait til tmr to upload pictures?

[charred parcel]

contact support, you shouldn't be needing to wait

you can contact support at https://help.openai.com

[neat path]

Anyone an idea how to upgrade from teams to pro?

[charred parcel]

you need to cancel Teams and start a new subscription to Pro. You can only upgrade from Plus to Pro not from Teams to Pro

[turbid vine]

since o1 is officially out for premium users, what is the promt limit ?

i have the plus subscription

[autumn sierra]

I am sending same image and prompt to 4o and 4o mini on playground. 4o shows 1200 input tokens but 4o mini shows 36k input tokens. if these numbers are correct 4o mini is more expensive. What am I missing here ?

[unborn coral]

how change the mail of the acc

[thorny crown]

i got this again today. im contacting support. ig they dont like u using o1 alot, even for smaller tasks ???

[trail ingot]

hi! Is chatgpt with canvas no longer available? I don't have the option in my model select dropdown although I'm a plus subscriber...

[icy maple]

Good grief, seriously? Is this everywhere? Or just me? Literally trying to make images of scfi/fantasy war machinery.

It seems to be on a per-conversation basis. Which is very odd. I have another conversation that I created (and used much more) using the exact same prompt, and that one worked fine just now.

[steep plinth]

hello I have a question I'm a bit of a novice in all this ia and the question is that I have several custom gpt to which I have loaded several documents in pdf or markdown as my knowledge base but I have reached its limit to me what I was interested in was to have a
a database which I could query all my documents internally from chat gpt which does it very well to bypass this limitation I use playground with a wizard and I have a vectorized database already or so my question is
Could I connect this database through the GPT actions with this GPT to consult in this Playground database the documents that I already have vectorized in it?
I know that what I am asking you may not make much sense since this database can be accessed from Playground but sometimes it is very convenient for me to connect from my specialized GPT from my mobile to this documentation and I would like to have as much data as possible and in Playground because there is no data limit for the vectorized database.