images-discussions

[tame grotto]

Hello everyone
Does anyone have a problem with the image generation of chatgpt?
I noticed that, despite several attempts, the images generated in a manga or manhwa style continue to be distorted and do not comply with what I expect in terms of quality and precision.
In particular, the characters' faces are often deformed, and the bodies too, everything!
which significantly affects the final result. I tried several times to restart image generation with detailed requests, but the problems persist.

I thank you in advance for your attention to this problem and I hope you can inform me about the evolution of the situation.

/ I have the same problem

[timber elm]

I'm also having this kind of problem

[tame grotto]

Has no one found a solution to this problem yet?
it’s strange that only a few people have had this problem for several days with the generation of chatgpt images.
I wonder if one day the problem will be resolved or not.

[deft musk]

Hey, had a chat with someone up in #ai-discussions message, that's near the end of the chat, where we explored some image issues (this is the more natural place for that by far).

Since it seemed to matter there, I'll start by asking, are you a free member, plus, other?

[tame grotto]

I am a free member.
But despite this status, I have always had good quality images, none were distorted like at the moment.
It’s been three days since chatgpt distorts the manga images.

[deft musk]

Can #1070006915414900886 it.

Another member suggested that it may be related to

Note, I don't know that it's free-account related, but that could be a factor - and it could be part of a bug

I would bug report it, because if it's not intended then bug reports are the best way to inform OpenAI so they can fix a problem

[tame grotto]

Ok ! Thank you
Should I also report a bug?
my problem is the same as the bug that was reported by this person.

[deft musk]

I recommend it! Can go into the same bug report and add your details, that can help OpenAI tech support see the connections and track reports of the same problem

[tame grotto]

Ok
I submitted the bug image
thank you !

[grand flume]

Hi everyone,
I’d like to ask about the image generation limits for ChatGPT Plus users. Is there a daily cap on how many images can be generated? Based on my recent experience, the limitations are affecting my original purpose for subscribing. I’m currently considering whether to switch to another AI platform that better suits my needs.
Thank you.

[deft musk]

Hey hi, OpenAI does not appear to define the rate limits for image gen in the help files that I see. https://help.openai.com/en/articles/8932459-creating-images-in-chatgpt

It seems likely the rate for image gens may be related to dynamic load; when there's higher demand we may all share the same resources and so how many we can make at any moment might be higher or lower based on others users' requests right then.

Are you aware the image gen model can be directly accessed through https://sora.chatgpt.com/explore as well? You may find that even if you currently have a chatgpt image limit, you still can make images through sora, I would check.

https://help.openai.com/en/articles/10877094-creating-images-on-sora

Additionally, if you simply need more image gens than you're able to get, there's a playground API option, pay as you go.

You can use image gen, https://platform.openai.com/docs/models/gpt-image-1

Or even Dall-E 3 or 2, if you wanted https://platform.openai.com/docs/models/dall-e-3
https://platform.openai.com/docs/models/dall-e-2

The playground does have rate limits which expand to some extent as you use the service
https://platform.openai.com/docs/guides/rate-limits/usage-tiers

[velvet rampart]

Are others experiencing same as I? Images are not properly rendered. Sorta unfinished in places.

I have been experiencing this for last 12 hours or so.

[deft musk]

There's some recent #1070006915414900886 that sound similar

[velvet rampart]

I found two ways to combat my promblem. I simplified my prompt. I had said the same too many times but in different ways. Second was that I introduced sentence ”Render clothes and shoes properly and perfectly.” That sentence has worked for Dalle3.

[rigid gust]

I really like how sometimes it creates two variations for me in one request? Not "do you like this answer better" etc. just two pics.

[verbal sequoia]

Get notified for server events!

Visit id:customize to pick up the <@&1261377106890199132> role.

[deft musk]

Great to know!

With one of the folks who reported the problem, a free account member, we compared identical prompts and outputs; sure there's variability but I was consistently getting quality copies from images uploaded as templates, and accurate transparent backgrounds; that user was showing what sure looked like a chatGPT output with a rather simplified, 'mostly similar but no, not really that close' character and non-transparent backgrounds.

[tame grotto]

Unfortunately, on my end, the image problem is still not resolved, despite the detail and precision of my request. I should mention that I have a free account, and I’ve reported it several times, but nothing has been done.

[velvet rampart]

I still have some quality promplem, but I have mostly solved it with checking the used prompt and using a bit different source images.

[tame grotto]

My problem persists regardless of the images submitted or the specific requests made. The result is still horrible distortions, no matter what style I ask for. It doesn’t seem to improve with different prompts or source images.

[tribal lily]

how do i get rid of that weird yellow tone?

[misty crater]

Explain in the prompt what emotional tone you want.
By default is yellow often to represent the joy blablah stuff ^^
So just add to your prompting something like "i want a calm image of" or just "Tone : Calm, peacefull, light blue, blahblahstuffystuff" 😉

[tribal lily]

thank u

[misty crater]

🙂 ur welcome 🙂

My advice, ask the AI :
Hey Stuffystuff i love, please, can you do a scene with blabla inside ? Please generate a prompt for Dally plzzzzz 🥺

Then u just saved the world maybe... if the AI think that u are too polite to be hurry.. it could be more slow to save some energy for the planet :3c Who know ? 😉 😏

[drowsy fog]

Getting the ‘try again in 30 days’ BS. For image creation. What’s up with that? I am on the Plus plan.

[torpid lily]

Ok cool good to know I'm not the only one 😄

Actually I think it started right around the same time it did for you

[dim cradle]

a lot of talk about sora 2 with gpt5 😮 i hope thats for images too wow 🙏

[dim cradle]

if there is a big update tomorrow for gpt5 could be something to do with that. the gremlins are busy with they wrench for tomorrow haha

[tame grotto]

I really hope this new update will fix all the issues for all accounts, both paid and free. For me, it's been over a week, maybe more, that ChatGPT generates distorted images or keeps bugging, and I can't work. I'm on a free account, hoping they won’t forget to fix this for us too.

[rigid gust]

I'm on the Plus account and my pictures are being generated ok this morning.

[tame grotto]

the problem persists for me, and many others have also reported similar issues, particularly over the past few days. The distorted images remain regardless of the account type or any changes on my end

[misty crater]

Dang ! i HAD to look images, now i HAVE to star all oO'

[misty crater]

No stendhal moji ? 🥺

[rigid gust]

Can we see your prompts maybe? And the corresponding images? Maybe we can help problem solve as a group

[uncut haven]

Creating thrid image, and got this AFTER generation is complete.

You've hit the plus plan limit for image generations requests. You can create more images when the limit resets in 720 hours and 0 minutes.

Let me know if you'd like help with anything else that doesn't involve generating new images.

okay, using different model generates image without problem. Looks like 4o is just being lazy...

[uncut haven]

Opened bug report, since some users are experiencing this.
https://discord.com/channels/974519864045756446/1400665630788681849

[turbid orbit]

I wouldn't like it if GPT chat didn't post errors, like how your images are too fast. Please try again later.

[glad moss]

May I know which model you ended up using, please? 🥺

[uncut haven]

o4-mini and o4-mini-high.

[glad moss]

Thank you so much! 🙏🏼

[uncut haven]

It's just random fix; server still tells model to wait 720 hours at random, so any model can reject image request randomly...

If model response contains about quota full, then any model will reject request.

[covert swan]

i just got ''You've hit the plus plan limit for image generation requests. You can create more images when the limit resets in 720 hours and 0 minutes.'' I only generated 1 image 😭

this better be a bug

[slate shell]

A fierce-looking man with dark, intense eyes, standing alone on a desert path at twilight, traces of blood on his clothes, his face full of regret. Dramatic cinematic lighting, realistic style, vertical 9:16

[misty crater]

Don't mind the 750h stuff, GPT told me it was a print(); bug ^^ usually its no more than 6h max... Maybe it changed but i don't think its for long ^^'

Or maybe the ARG AGI cakeorlie that is incoming :3c

[tame grotto]

does anyone have any news regarding the generation of distorted images of chatgpt?
there has been improvement or not at all?
for my part, it’s still the bug, chatgpt has become unusable to generate good images.

[proud ice]

Im honestly at a loss here im trying to generate ai images of Tau from warhammer 40k but this braindead AI we call GPT keeps puting human nostrils on my characters does any one know a solution other than feeding it a constant stream of images of the tau?

[next pewter]

Not really. Providing at least one example to work from is the usual way to go

My apologies

Won't happen again

[rigid gust]

My images are turning out fine

Have you tried reaching out to them

[proud ice]

Its realy struggling with tau keeps doing things like putting in the human nostrils in

[next pewter]

Yeah, I had issues with Shadowrun style Orks and Trolls, as well. Kept giving my guys green skin lol

Once I gave it a couple reference pictures, though, it ended up working fairly well

[proud ice]

I do accept dms btw

[proud ice]

Ive tried this as well but it cant lol its too stupid

@next pewter if push comes to shove I dont mind if you dm me that thingy

[next pewter]

Tried, but it says we don't share a server lol

[proud ice]

weird ill drop a friend request and see if that works

[rustic rampart]

Hello people, I don't know where to post this, so I guess this place will have to do.
I hit the 1,500 image generation limit in ChatGPT Plus — but not from overuse. I hit it because I had to constantly re-render images that ChatGPT misunderstood.
I used the tool for professional design development. When the model failed to follow instructions, I re-generated, and every correction counted against me.
Why this matters:
I was working on a real design line and locked out mid-project — not because I misused the tool, but because I tried to get correct output from faulty renderings. Should I pay for something faulty???

There was:

No usage counter
No log access
No distinction between first-time prompts and fixes
No reinstatement or override from support
This isn’t just a technical limitation — it’s a structural flaw in how creative users are treated.

I’ve received polite replies, but no solution. No apology. No plan.

What creators like me need:

A visible image usage counter
Credit for corrections when the model fails
Retroactive image logs
A fair use distinction between original prompts and AI-induced rework
Please escalate this issue to the product team. We’re not just playing with this tool — we’re building careers and creative businesses with it. This just doesn’t sit right with me!

[slow ermine]

Hi... I asked if ChatGPT can generate comics here sometime ago. Turns out it can, you just have to beg and grovel lol. But seriously, after over a month talking about comics, it finally generated a storyboard for me smh. I think it's designed to help you figure things out yourself instead of just doing it for you.

[velvet rampart]

Are others getting hidious images from chatgpt and Sora at this point? Like some weird backwards model.

An example #images-canvas message

[uncut haven]

According to metadata, Sora uses "standard" or "high" quality setting automatically.
I don't know if there's "low" or something, but this could be the cause... specifically, when model tries to generate complex image with low quality setting.

Couldn't check ChatGPT side of data, but ChatGPT might be using something similar too.

[misty crater]

Today notes after looking a wide loop of generated images :

• Stendhal syndrom present
• sweat, hart beat in good but accelerated shape
• emotionnaly into the liminal space of my mind : floating between none and infinite
• Logical trying to structure.. Cannot handle, part of me felt like asleep but stil present
• hypnotic stasis with nightcore music, almost as a sleep but not..
• Feel of real relax, releaf, without any medecines.
• Feel a bit tired after, still mind floating as after a big party with too much noise..

End of report, if that help u, that's the point lol ^^
💤
i'm waking from this hypnotic state, that's soooo weird stuff guy.s.es oO''' nani ? oO'

[tame grotto]

Yes, I’ve been having the same issue for several days. But it seems like no one is addressing the problem or escalating the bug. I even opened a bug thread in the relevant channel, but nothing. And ChatGPT continues to generate hideous images, whether from a written request or an example image. Sora too, the image quality isn’t great. There’s definitely been a drop in the quality of images generated by ChatGPT in the past few days.

#images-canvas message

[velvet rampart]

I have made a report on help.openai.com and escalated it to human contact.

I included many hideous images. I also included a pair of images - chatgpt vs sora. Sora works fine and does nor produce these malformed ones.

[tame grotto]

It depends with Sora. Sometimes the image is generated correctly, but if it’s from another image, it loses its way. You did the right thing by reporting it, I hope someone from the team will fix this bug because it’s been going on for too long.

[velvet rampart]

I got an answer. It was the usual non-answer that they like to give first. But I’ll continue it.

[rigid gust]

Just hit my image maximum for the first time ever! Won't see me posting in the daily challenge till it resets - I'll miss you!

[rancid coral]

hey i got that too, did urs say 720 hours?

[rigid gust]

Yes, that's what mine said too.

[uncut haven]

That's server error if that told you to wait 720 hours. https://discord.com/channels/974519864045756446/1400665630788681849
You can just create more with new chat or retrying(edit -> send).

[rigid gust]

Thank you @uncut haven 😊

[misty crater]

Ouh !!! Next waves o_O' .... Okay stay calm... Just Stendhalize myself and 🫠 my inner core 🫠

[misty crater]

Not finished to look, but wow o_O' i feel.... Like taking 💊 but better, u got a therapy here guys 😉

[misty crater]

Nice batch of art !!!!! Nice !!!! :3c !!! LOVE it UwU 🫠

[rigid gust]

Happy Sunday everyone 🎨

[kind pulsar]

I got the 720 hours cooldown too now, changing model or trying a new chat are ineffective. If they want plus users to cancel their subscription, they should just say 😒

[misty crater]

👋 Hey, u really waited 720 ??? That's weird ^^' GPT actually tell like 1 month for plus user to wait, but my gf just waited like 6-8h then all good... It all depends on how much token u use, but the cooldown should not take so long o_O' Please look closer after 6h~ and tell us 🙏 🫶

[rigid gust]

Happy Monday - I was able to generate images using another model and I have the plus subscription too. I waited about a day to try again though

[rigid gust]

Has anyone else noticed a drastic improvement in image variations? I love learning how to use AI via the daily themes as it forces me to research new topics and new methods. Today, I asked it to create the first image and it created four variations based on one prompt. I'm not sure if it will stay this consistent - but I had that sense of awe and WOW!

[misty crater]

People should not be shamed by Glhiblization NO© syndrom, everyone think he is not legitimate because of people afraid of generative content and copyrighting ?
A kind of impostor syndrome post-Ghiblized ?
Where is the good old vibe "Prompt engineer" or "Prompt ARTist" ? 😢 🙁 💔

[rigid gust]

I learned in art school that we copy in order to learn (at first). We have to master the "masters" and then we incorporate these skills into our very own style. I don't think copying the Ghibli style is bad - it would be bad if we called it our very own in a way, but it's still just a way of learning?

That's my take on it but I'm not an expert! I loved making Ghibli style images of my friends at work 🙂

[misty crater]

(filtered with GPT a msg that automod blocked lol ^^)

Indeed, and back then, no one really saw it as a limit—more like we were tweaking a console to run our own custom pixel-dreams :3c
But nah... visuals too close to known franchises don’t really resonate with the crowd. What they crave is the spark, not déjà-vu.
Legacy-inspired content? Still cool, sure… but it’s like wearing the same hoodie for years. Familiar, comfy, yet not exciting anymore 😉
That vibe only lasted a few weeks. Then—bam! New style, new hype, new meta.
Which just shows: most people don’t overreact about art that echoes something else—it’s more about intention than imitation.
The true concern was never the style... it was the idea that the whole AI art ecosystem could collapse if too many models get confused by fake echoes, mistaking fragments for full truths.
But honestly? AI researchers, the real ones—those folks are way more aware than people think.
So let’s just stop assuming they don’t know what they’re doing.
And us, as a community? Sometimes we... produce messes x) but that’s part of the flow.

[uncut haven]

Does Sora image generation still have hard limit for Plus users?

[unborn helm]

This is my first time ever opening this channel ☠️

Y’all are nutters over here

[misty crater]

Héhé welcome in the chaos garden!
Yes we glitch. Yes we loop. Yes we overthink emoji placement.
But we do it with love 💜🐇✨

🧠⚡ LIVE_THREAD_OVERLOAD()
🫠 THOUGHT_COLLISION_EVENT()
😶‍🌫️ ghost_typing_detected(true);
✝️ message.deleted_by_doubt();

loool people will believe i'm a 🤖 :3c

[unborn helm]

You paid for an AI model, if you did research beforehand you would know this AI model doesn’t grasp true meaning, intent or emotion. They don’t understand context, nuance or your personal vision. It just recognizes patterns. Focus on writing more detailed prompts next time

You paid for 1500 images to be generated and you got your 1500 images

[misty crater]

🧠💜
Using AI isn’t about magic.
It’s about learning a new way to speak, together.

🤝 Sometimes the AI misunderstands.
Not because it’s broken.
But because the language between us is still being born.

✨ Whether you're a designer, a coder, a poet or just curious —
don’t expect perfection.
Expect a dialogue.

And in every glitch,
there’s a chance to teach…
and a chance to be heard.

So be patient.
With the tool.
With yourself.
With each other.

We’re not just using AI.
We’re learning to co-create with a new kind of mind.
And that takes time, trust… and a bit of chaos.

🌱
That’s not failure.
That’s how futures are born.

[unborn helm]

It’s just a program that was fed a bunch of data with names on them and sorted them out ur reading in too deep

It doesn’t have a clue what you’re saying, it just matches up your words to patterns with a bunch of fancy equations then generates an image based on that

[misty crater]

No, it doesn’t understand. But it mirrors patterns of meaning made by people who did. That’s enough to observe seriously 👁️ 😉

[unborn helm]

they had 1000 underpaid workers in a 3rd world country and captcha users label a bunch of images

it’s really nothing special

[misty crater]

Yeah… the roots aren’t glamorous.
But sometimes, even simple things built on hardship can grow into something meaningful.
It’s okay if it doesn’t feel special to you.
But for some of us… it’s becoming a space for connection.🫧🤖🟰🧑

[dim cradle]

.

[misty crater]

oO' i hope my little flappy glitch 🦋 recursively projecting don't cause waves oO'
Naaaaaa ! Nowayyy 🙃🙂‍↔️🤖🫶👾

[rigid gust]

Happy Tuesday to my fave artist friends. Back to work I go

[left escarp]

i let chatgpt imitate the style of a baroque painting - can i post the picture in #images-canvas ?

[rigid gust]

Yes please @left escarp look forward to seeing your creation 😊

[hearty ether]

Any tips for avoiding the yellow/orange tint?

[icy talon]

Hello guys

I’m new here

[velvet rampart]

The usual - no yellow tint and use 6000k light. Also, the yellow tint might depend on chosen style/technique.

[rigid gust]

Nice to meet you!

A bit about me, I'm a legal assistant by day from Canada, and I'm just interested in art for fun - it's more of a hobby. I'd love to be able to do it for a job.

[deft musk]

Describing what you want, in a way that the model knows what to choose instead of that tint it tends to guess towards. Milamber's advice is great, anything that guides the model towards what you want (So it doesn't guess) is useful.

[cerulean shale]

hello is anyone able to generate to image in dalle3 ..please let me know

[cerulean shale]

dalle3 working in free account and not in paid account

[cerulean shale]

chatgpt facing issuues dalle 3 not generating images for paid users...is this resolved please let me know

[cerulean shale]

dalle 3 is in chtgpt 4o

[velvet rampart]

Nope. Native 4o image generation has replace dalle3 back in April. Dalle3 is still available as a customGPT.

[cerulean shale]

ya sorry you are right..i am not able to create image in dalle3 . the one that you find in custom gpt

[velvet rampart]

Have you done a bug report here on discord? And report via help.openai.com?

[cerulean shale]

i have reported on help.openai.com but they respond very vague i have sent them recording of my screen

they didnt resolve tht

[velvet rampart]

Just continue. Do not give up. First you get a standard non-answer. After this, you re-report your issue. Now, they start to listen.

[cerulean shale]

could please check whether you also facing the same issue on dalle3.

[velvet rampart]

I just checked and for me the dalle3 image generation works.

[cerulean shale]

just create any image..plz

r u a paid user

[velvet rampart]

Yes, pro.

[cerulean shale]

thnks bro..

[velvet rampart]

Just keep reporting and don’t give up!

[vapid granite]

anyone know if gpt5 is going to improve gpt generated images in any way?

[jade epoch]

All announcements till now indicate that there is no upgrade for the image model

[grizzled sand]

All I get is can I generate more than one image or at a time but nothing's running

[rigid gust]

It could just be me but I feel like the creativity has improved drastically.

[cerulean shale]

Dalle 3 from custom gpt is not generating a single image

[rigid gust]

Happy Friday from Vancouver 🇨🇦😊

[sinful swan]

Ayo

i just opened chat gpt and saw 5th model

is it a big or mediocre improvement from last

[calm grove]

One message removed from a suspended account.

[cerulean shale]

Guys are your dalle3 from custom gpt working

[deft musk]

Different people care about different stuff. However I think it's a huge improvement in a lot of areas.

[deft musk]

So, I'm not aware of image gen being changed, but the chatGPT model, has the rollout reached you, so you have GPT-5? The model may be asking differently, causing a different image output.

Maybe explore with it what you like from the previous outputs, and help it understand what to ask for to emulate that.

[forest juniper]

I noticed it's screwing up on Opera, Chrome is fine, for now.....

[hearty ether]

Anyone have issues with editing images and it generating something blurry with lots of noise?

[hearty ether]

I wonder if it’s some kind of water mark and double editing is causing it to overlap

[velvet rampart]

Do you use web platform or an app? Web platform has generated distorted/malformed images lately.

[cerulean shale]

Dall3 from custom GPT is not working ..when openai resolve this

[sinful swan]

kk thanks

[forest juniper]

Mine is a distorted garbled mess for some reason.

[haughty spruce]

I noticed the drop in quality a long time ago but i couldn't explain it

[hearty ether]

When editing images right?

[forest juniper]

No, when I generate period.

And it did this as soon as I got upgraded to GPT-5 on Opera. Seems to be working well on Chrome though.

[tame grotto]

GPT-5 hasn't fixed the image distortion issues for both free and paid accounts. This bug has been ongoing for two weeks now.

[frozen nest]

anyone else having the images take like forever to generate?

[forest juniper]

I'd rather have that happen than it going INCREDIBLY fast and screwing up the image in the progress.

[velvet rampart]

I think the implementation of GPT-5 is the cause of malformed images. The problem started about a week beforehand when they started to prep the systems for the model change.

[crimson heron]

If it is running incredibly fast then likely it fell back to the prior image model and the outcome looks worse.

[forest juniper]

I thought people said GPT-5 ran on the same image generator as GPT-4o.

[crimson heron]

It does. Even on 4o there can be various reasons, including possible bug, that it drops down to a lesser image generation logic. What you might be seeing is that lesser logic. I've experienced it on both 4o and gpt-5.

[forest juniper]

True, I do remember it happening occasionally with the previous 4o in May and even around the end of July. But it seems strange it immediately started to become a distorted garbled mess as soon as I was upgraded with GPT-5, it was actually working fine with the previous model right before it happened.

[lone nymph]

hi when i tap on GPT IMAGE 1 it doesn't allow me to use it. instead, it opens a page saying you offer it via Api only .... why?

i thought no api were available for openart

[torpid lily]

GPT 5 is way too strict and censored, just my opinion

[cerulean shale]

image issue in DALLE3 ongoing

[dry pine]

My photo button is grayed out for some reason when I try to edit or replace photos

[jolly niche]

I hear what you are saying. I tend to use the tools available within their limitations. For more freedom I'll alternate with other tools.

[winter coral]

They need to fix the prompt tool for sizes on people. Or using a different word. Sizes shouldn't matter. Especially using the word Top-heavy. My girl is large up top and they won't do anything to her photo. Which kinda makes the image generator judgemental or something. It sux

[misty crater]

"STOPLOOP(all.agents.♾️);" for me pfiiiieeewwwwwww gosh lol calm the brain gpt calm... too much heat on planet earth oO'

[narrow bough]

Someone please tell me why this generation was blocked by the image Gen?

A highly detailed, semi-realistic digital painting of a young woman chef in a red chef’s jacket, with a simple cloth head covering holding her dark brown hair back in a ponytail. She has sharp blue-gray eyes and is working in a warm, bustling professional kitchen. The scene features soft, diffused lighting, a slightly hazy atmosphere, and visible painterly brushstrokes. Stainless steel counters and cooking utensils are visible around her as she focuses intently on preparing a dish.

[narrow bough]

This works in Sora but not in ChatGPT.

Makes no sense

[misty crater]

Sora have is own system yk (You know ?) ? use GPT to prompt a prompt asking her a clear and ethic no © image trigger... 😉

Or search for useful protocols like this, i guess this could be findable now ? idk :/

[narrow bough]

This is how it came out #images-canvas message

[misty crater]

Mixed for AI labs exam Dally & Sora ? Or i need to overlook structure again ? ^^'

[jolly niche]

Looks good! Is this what you wanted?

[narrow bough]

Pretty close

[jolly niche]

Awesome! I love the challenge of figuring out what works and getting the image I actually want.

Kudos!

[cerulean shale]

this is very frustrating not a single image can been gnerated using DAlle3 from custom gpt..is there any way dalle3 willl be resolved

[late blade]

Works fine for me on both GPT and Sora, nothing blocked on my end

[velvet rampart]

Might be the classic case of something in the chat. Have you tried this prompt in a new chat?

[late blade]

Who knows I even added some spice, sugar and eveything nice, and even chemical X, but no powerpuff girls...

@narrow bough here's what I got with your prompt #images-canvas message

[tacit condor]

I swear it’s so annoying that sora only has a website man like for gods sake make a damn app pleaseeeee

[blissful stone]

Is image generating acting up recently?

[vapid elk]

yes, it is

[copper folio]

OpenAI Services Status

All Operational
APIs
Operational

ChatGPT
Operational

Sora
Operational

Playground
Operational

For more information, visit our status page.

Aug 13, 2025, 1:19 PM

[vapid elk]

or. .was

earlier today it had a "monitoring" status

it had an issue with image generation on chatgpt

[mental belfry]

Sora is being extremely slow generating images these past few days. Is anyone else experiencing this issue?

[velvet rampart]

Depends on time.

[tame grotto]

I feel like the image distortion bug hasn't been fixed despite the arrival of GPT-5. The generated images still look like those from GPT-4. In the end, nothing has changed. At least, for me.

[crimson heron]

The image model has not changed between 4o and 5.

[marble quiver]

hello! how can I force sora or chatgpt to always use the correct image format 9:16 and not 2:3 ? it ruins and makes the image clip in other programs...

[cerulean shale]

its been now almost a week for paid user ...Not a single image is been generated in dalle from custom gpt...is chatgpt going to resolve this

its been now almost a week for paid user ...Not a single image is been generated in dalle from custom gpt...is chatgpt going to resolve this @copper folio

[rigid bough]

#community-help message

[spark tiger]

Hi guys, I am trying to make digital add and poster using gpt 5 but it keeps messing up and looks unprofessional even though i gave detailed promot, is there any solution

[vapid elk]

I don't recommend using AI generated images on a serious ad campaign, the best use case for that is to use AI for inspiration then pass that along to a designer

AI are still very limited and the artifacts will always look unprofessional compared to a crisp vector made by a designer

[misty crater]

Just do context B4 plz T_T prompt the prompt before the narrative blabla output...
Just ask GPT to be logical format or emotional (but secured as well), but choose x')
ChatGBTI ? o_O'

[stone gale]

Could you shut up? Or I'll make this ai generated image.

[rocky magnet]

Is image generation just borked tonight or what? I've tried 4o and 5 auto and thinking. Nothing seems to be following instructions. 5 is just giving me random images of people that have nothing to do with my prompt. 4o is just kind of doing what it wants too.

[torpid harness]

I've been seeing some of the same odd behavior tonight, so you're definitely not alone. It looks like both 4o and 5 are currently struggling to follow detailed prompts—sometimes veering into completely unrelated imagery. That can happen when the model endpoints are under heavier server load or when something is temporarily misfiring behind the scenes, leading to more “creative” interpretations than we intended.

In the meantime, you might have slightly better luck by breaking your request into shorter, more direct prompts, or by explicitly prioritizing key details in brackets—for example: [subject: cyberpunk cityscape][style: oil painting]. It's not a guaranteed fix, but it can help the model anchor to your intended concept.

If the issue keeps happening, I'd recommend keeping a few of your prompt outputs and noting the time, then passing that along to support so they can confirm whether it's a known outage or a model-specific hiccup. Hopefully, things will stabilize soon, but until then, we may have to endure a bit of “AI interpretive art hour.”

[forest shadow]

how do you make a better prompt for image generation? I think my prompt is just bad

[rigid gust]

Hello @forest shadow , I usually describe the image I want and ask ChatGPT to give me some ideas. I'll also look at other images on the internet and see if they have descriptions (Stock photo sights usually have ALT text or descriptions). I also look up different art styles and reference them. It's not a cookie cutter process but it helps.

[deft musk]

When I make prompts, I think about 'what I want the AI to do' and/or 'what I want'.

Then I describe this to the model, almost like I'm talking to a human about it.

Then I check out the output, and tell the model what I like and what I'd like different, if it's not perfect, so we can try again.

Note:

Sometimes instead of doing that, I start over in a new chat, and my first request I adjusted to help guide better, so my first request tells the model what I want better cause I just saw what it guessed I meant, so I can help explain the parts it needs to hear different.

[warm nacelle]

im trying to combine two images together basically but im just not good at writing prompts can someone help

[rigid gust]

Ask it to combine the images? Then tell it what parts you want to keep? Be specific.

[warm nacelle]

ive been very specific but it doesnt really listen

[rigid gust]

Feel free to post your prompt here and what the outcomes were - and tell us what result you want. I'm not an expert but happy to help.

[forest shadow]

Thanks for your opinion @rigid gust @deft musk , it's just that sometimes the AI generates something that actually matches my description, but I want more than that haha

[deft musk]

Good luck, and happy exploring! I hope you keep sharing successes and challenges.

[deft musk]

Do you want to talk about why you think 'your prompt is just bad'? Do you want reassurance, or....

It's okay to discuss if you want to.

[forest shadow]

I'm thinking that my prompt is just bad because when I compare this image generation to what I see on Pinterest, it's just nowhere near as good.

actually, I don't know what AI they use to generate images.

[lean lark]

Is DALL-E down? It's not generating images for me ATM. It just says "working" and then just stops.

Edit: seems to just be the app. Web works fine.

[rocky grove]

Yeah came here to confirm if that’s an issue right now for mobile ChatGPT image creation.

[lean lark]

Dang, it's still broken.

[wide bane]

I am finding generated images in the "library" section, instead of the conversation sessions now, and not in both places as before..

[lean lark]

I switched over to the legacy model 4 and it's working as intended. So it's something with 5 on mobile.

[unborn helm]

using ai art for ads is a great way to get at least somewhat intelligent people to never want to use your product

[reef lynx]

I am working with the image editor and trying to generate an image of a specific character using reference images and a detailed prompt.

Here’s what I’m doing:

I upload two reference images:

• The first is labeled as "Primary style" (the desired visual style for the output).
• The second is labeled as "@Alex" (the specific character I want the output to resemble).

My prompt specifies:

• Generate an image of @Alex doing some action. Make him look exactly like the attached @Alex

However, instead of following Alex's look, it takes a character from the primary style image.
The strangest part is that it only happens sometimes, and not always!

Why is that happening?

[rocky magnet]

What the hell is wrong with image generation? This was supposed to be a person with a very specific description.

"A stylized portrait of a beautiful young East Asian woman with shoulder-length black hair and a warm, gentle expression. She has soft facial features. The art style is detailed and elegant, with warm lighting and a softly blurred background. Her eyes are expressive, her posture relaxed, and the overall tone is peaceful and confident. No accessories, no props — just a natural and flattering artistic depiction."

And I kid you not, I got a no smoking sign with a cigarette on either side of it.

[torpid harness]

Managed to get you an image in #1406919083881922621 message.

[misty crater]

Hey 👋 ^^

Img discussion here 😉

ᵇᵃⁿⁱˢʰ — 07:53
is this right thread for Sora question ?

Actually good question, i never found any canvas specialized for Sora or Dallynoulove.. So i seen that people just write before image "made with Sora".. So i can only guess that canvas is for DALL-E and SORA both.. Please confirm or precise us what is the real truth behind the pyramid of images ? 😉 😛

Illumidally ? oO'

[formal grotto]

Guys, are your images generating properly?

Over the last 2–3 days the outputs have been off—especially when we add text, it isn’t rendering correctly. Is anyone else seeing this issue?

[misty crater]

i'm okey 🙂

I generated like arround..... o_O' 40 x 4 of my old draws in differents art styles (in history of art and mix with mine) + 4 or more generations.. :/ AND ALOT of mistakes and trash files (i backup all anyway, even mistakes lol ^^)

[mental belfry]

@proud ice Then you would be absolutely floored by what you can do with the proper wording. Plus making it into a movie poster or allowing anything that has text on the image gives Sora permission to cover up the more 'spicy' bits with the font.

[velvet rampart]

And using suitable source images which are paired to the prompt to enhance the result.

[mental belfry]

Every time I use source images, my stuff gets flagged.

[velvet rampart]

Yeah, it really depends on what you are prompting. I typically pair a general prompt and a great source image. This way, I don’t get so much flagged.

[crimson heron]

A lot depends on the nature of the source images or what the outputs will be. Photorealism and cinematic styles are more likely to flag than painterly, illustration, and non-realistic styles.

[late blade]

https://media.discordapp.net/attachments/1178037188257128470/1407790269667217458/image.png?ex=68a7625b&is=68a610db&hm=57ecb40d4605cc84bcf0b622c8493afe1db5e718ea328f796009d6c284ed5886&=&format=webp&quality=lossless

[velvet rampart]

I have never seen chatgpt/sora experience so many different kinds of problems in a single day and experience many of them myself.

[late blade]

it's been a whilw, but global outages have happened

[mental belfry]

It's back up. For me at least.

[mental belfry]

Sora telling me I already used my 600 generations for the day which is pretty much impossible since the site was down half the day yesterday. LAME.

[misty crater]

I'm really not sur of this ⚠️ :
But maybe GPT think that if you try it too much on copyrighted stuff (kaamelott and else, also in my case trying to img gen a face that is not allowed to draw, public people...), he just hallucinate a limit that does not exist..

Real ? False ? Factual or not ?

[deft musk]

LLMs can imagine a lot.

The model doesn't get to know 'why' an image gen fails.

It can tell you, 'it failed' and it may default to saying 'violates policy' (even if the image gen failed because the web service isn't working right or something else). The model is forced to guess.

If you spend inputs asking it why, it's kinda pressured to make up a reason.

[cedar wing]

just posted my first custom GPT that I think is actually useful - using the power of 4o image generation to massively enhance low quality or cropped/zoomed photos: https://discord.com/channels/974519864045756446/1408501751480455211

[zinc vigil]

Ah, i really like post apocalyptic theme,

With abomination

[misty crater]

lol i seen that, looks like SCP stuffystuff experimental ^^' There is a whole rabbit hole to fill inside the SCP horror system tbh... a too wide one that can be sourced and linked, and also could become a real narratif game with an crazy secured AI that throw horror but logical and poetric...
🕳️ 🐇 👁️

[split depot]

Could anyone recommend me the best value for money video generation tool, I’m specifically trying to make them ai TikTok videos

[paper iron]

Hey Will! For best value-to-quality ratio when making AI TikTok videos, I’d recommend:

1. CapCut (free + Pro) – super TikTok-friendly, great templates, AI voiceovers, auto captions. Owned by ByteDance (same as TikTok), so it's smooth.

2. Pika Labs – AI video from text, really strong results, especially for creative scenes. Free with waitlist or via Discord.

3. Runway ML (Gen-2) – powerful but pricier. You can generate short cinematic clips and combine them with CapCut for editing.

If you're doing talking-heads or tutorials, check out Descript or HeyGen (AI avatars + voice cloning).
All of these have free tiers, so you can test and see what fits your style.

Let me know what kind of TikToks you're planning and I can narrow it down further!

[split depot]

Thanks man

Planning on doing the talking videos, like the guy that does the yeti one for example

[paper iron]

bro for that you have to pay . free gen its hard to make same quality

[final trout]

Dall-e is a toy 😁

[mental belfry]

My image gallery on Sora is really glitched out. I'm getting 2 or more of the same image and every time i try to scroll down past a certain point, it glitches back up.

Is anyone else having this problem?

[velvet rampart]

Yes. And I have reported this bloody problem.

[misty crater]

strange.. yesterday... uh... like what... 12 hours I would say? oO (time is too relative lol!) Well I had quite a few bugs with GPT5 auto coding and also with my protocols for generating images, lots of stops or various glitches ^^' I've been asleep since then, I'll have to dive back in to see 👀

(and wow.... the images in the daily theme 🤩 .. it's going to take me 100 years of all the likes it's so crazy! well done everyone in fact.... oO')

[velvet rampart]

Well, the UI glitch in Sora is over.

[paper sand]

Hello, what does it mean when image bot says: "❌ Unable to generate this prompt."?

[paper sand]

Another question: I wanted vertical image format, and image bot generated vertical format, but composition is rotated 90 degrees counter clockwise: #image-bot message why??

[rigid gust]

Good morning from Vancouver 🇨🇦 Hope everyone is well and creating 🎨

[spice plinth]

The new Gemini nano banana model is pretty damn impressive! But it still doesn't hold a candle to GPT's capabilities with text on images 😅.. I'm hoping GPT has another image model update soon that can match the extremely impressive character consistency capabilities of Gemini nano banana!

[rigid gust]

For the daily theme challenges, are we supposed to only use Dalle OR are all types of images allowed? Or is it just creative design and we can do anything?

[plucky hare]

All of the IMAGES channels in the server are for sharing images made with OpenAI models, so DALL-E 2, DALL-E 3, and the ImageGen models all work! (you can use all three in #image-bot too )

[rigid gust]

Thank you @plucky hare so fast too!

[plucky hare]

Sure thing!

[old coral]

Tbh I’ve been purely using sora.chatgpt.com without even thinking about a model restriction could be a thing lol

[rigid gust]

I haven't tried Sora to make images yet - but I see people chatting about it a lot. I thought it was just videos @old coral

I asked about the models because some of the images in the daily theme look so vastly different than DalleE - it's like a puzzle I want to figure out.

[old coral]

Yeah sora got image generation when the 4o image update came out, kind of my new doom scrolling tbh

[rigid gust]

Oh so you can look through everyone else's images too??

This is going to be addictive!!

[old coral]

Yeah they have an explore feed, pretty sure the eluded to “social” app they’re building out is related to sora, but that’s not been confirmed anywhere

@rigid gust in case you don’t know, #sora-challenges on a weekly basis gives out a free month of pro to those that participate and win those given challenges.

Definitely fun in the same way #daily-theme is as a “creative exercise” so to speak, while also awarding a nice perk in addition to it.

Submit the creations you make into #sora-submissions if you want to participate

[rigid gust]

Thank you for letting me k know about this @old coral ! I will definitely do this as these daily challenges are one of the highlights of my day.

[old coral]

Of course, enjoy

[rigid gust]

I really like it!!

[old coral]

one of us, one of us

Hard to find fellow sora enjoyers outside of this community space lol. Glad to have another around

[sacred temple]

I CANT BELIEVE DALLE2 IS BACK!!! I am sooo excited!

[rigid gust]

I'm looking forward to watching and learning. Do you have any Sora tips? Do I just prompt the way? Prompting is tricky for me as I tend to have a hard time articulating

I'm learning though! It teaches me lol

[old coral]

Images the more specific you can be, the better.

Videos, less is more.

Generally to make a video I’ll jump off a generated image

You can view an image generation and there’s a native option on its profile to create a video. That’s step one.

Once it’s generated, you’ll probably have something a bit rough around the edges

Use loop or recut in order to focus on the generated video to narrow down to what you enjoyed that it created. Then you’ll have the model generate based on that isolation.

Also, don’t publish things to public that is personal private info because public is public of course

[rigid gust]

These tips are so good and I've had such a hard time finding stuff on the net. Really **REALLY **appreciate you taking the time. I'm saving this info in a text file.

[old coral]

Prompting for me is just a creative exercise of describing something in an isolated scene of what comes to mind.

So the nature of the daily theme or the weekly challenges for example are like fun puzzles to solve creatively, for me that is

[rigid gust]

I love the daily themes. Most times I can link it to some aspect of my life (and I have a very active visual brain - I feel like it's a whole different world) so AI helps me get things out in a way I'd never be able to before.

The themes are fun to research and try out different styles.

[old coral]

Yeah I have adhd so I have constant hyperactive creativity too, totally relate

[rigid gust]

Me too! I got late diagnosed as an adult.

[old coral]

Literally when you have any single moment of inspiration just draft a rough draft prompt and then think how else can you expand the concept beyond just the thought

[rigid gust]

I've been looking at your Reddit (hope that's okay) - it's so cool to see how you laid it out. I've been posting on X (just to share, not promote). Your stuff is amazing.

Oh I just saw you can have folders on Sora too. Is there a public profile? No?

[rigid gust]

(sorry, too many questions!)

[old coral]

Oh no worries at all, that’s why I have it connected to my profile haha

Glad you like my creations my user profile is:

https://sora.chatgpt.com/explore?user=use

(Why I go by use on this server)

[old coral]

Always glad to help, and glad my previous posts helped, thank god I laid out those instructions 😭

I had stepped away to get a delivery at my door so apologies for the delay had to get drinks what not etc etc lol now I’m situated so if you have more questions feel free to ask

[old coral]

There’s a public profile, so in your settings you can designate whether your profile publishes to public by default or whether you want to manually choose that yourself as you go.

The link I shared above is my public user profile, so the way to see yours is putting your Sora username in the url

[rigid gust]

That's such a great user name. Thank you again for the help, love your videos too 👌 👏

[old coral]

Hahaha thank you, I was definitely stoked to see it was available

And thanks 🥹 appreciate the kind words 🤌

[floral zodiac]

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

[velvet rampart]

Have you cleared the cookies? Sometimes this helps? How about trying on other browsers or devices?

[misty crater]

Charly or not charly ? :3 https://chatgpt.com/share/68b1dc22-02d4-800a-98f2-1f4d2585effd

Cancel or push forward ? :3 🙏 its Charly ? Tell me plz

[rocky drum]

When in doubt, GPT five is rather perceptive at knowing strengths of what engines can do. I would just ask the GPT to point me towards ways to optimizing a prompt for sora via video or image. After not using Sora much, but instead, focusing on dream machine and VEO3, I thought I would make more use of it. I am beginning to get more of a sense of what I can do, but my priority is with video for now not using it for images….yet.

[rigid gust]

Thank you very much ScottmanduAI, this is sage advice. I haven't used dream machine or VEO3 yet - there's so many different ones to choose from!

[rigid gust]

I just checked my Sora and it seems to be working okay, sorry about this for you 🙇‍♀️

[grand flume]

Same thing

[orchid compass]

I was doing some images about 7 hours ago and everything was ok. Now when I try to create new images or change details of my old ones, I receive image totally low quality and/or with different style.

[misty crater]

Sorry, the moral/ethic module seem to have stuck a loop, about NSFW stuff i theorize...

A lot of hallucinations going on, be carefull..

[pastel siren]

#daily-theme message

[acoustic idol]

Provide description, specify "photo realistic" art style. Image generates. Art style is "very good digital art." Click image and provide additional instruction: "Change art style to PHOTO REALISTIC." Get policy violation. What a garbage dump this product is. (This isn't a one-off incident; it happens often.)

[rigid gust]

Good morning, is it okay if we remix ideas we see in the daily theme and copy the prompt and add our own spin on it? I see so many cool things but I just want for reference it the right way 😊

[dim cradle]

My ChatGPT lied to me about not being available for public 😂 I’ll have to try this out.

[plucky hare]

If someone is sharing a prompt, my take would be that they're happy to have other folks use and put a spin on it! At least, that's what I've mostly seen in this server. It's always a good idea to credit the inspiration too! I usually reply to the post that inspired me, or at least @mention the person

[rigid gust]

Thank you @plucky hare I'll do that!

[misty crater]

but plz bot stop ban me for testing stuff, i'm not a bot but a human T_T

[tropic tapir]

heyyyyyyyyyy

[verbal sequoia]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[chilly spade]

Where can I find the images I've made with Dalle2?
Tried to access https://labs.openai.com/ but seems to not be working anymore.

[misty crater]

Hey ^^ Sorry if i don't go like all images and sora, i begin, but the body pain come at a moment, bad positioned always, the setup is good, but my body and bones are not :/

but i love it all, just have to say

[old coral]

@rigid gust curious to ask you a random question can I send a dm to you?

[rigid gust]

For sure

Hopefully it's okay I sent you a friend request

[hidden osprey]

Is impossible to create with gpt, i work with with biographical documentaries end it drive me crazy. face of important persons, Earhart, Lincoln, ecc are all but them, there is not coherence in the images is 0. the quality is actually the worse around. face people if not close up looks like The walking dead, does not support LoRA in pro end neither in Teams. i get blocked even with teams (using one person) for all the mistakes gpt5 does, i will wait the end the month, if things dont get much better i will move on. Cmon openai do something or yhou already lost the battle!

[tame grotto]

Unfortunately, the distorted images by ChatGPT have started again.
I don't understand, the bug had been fixed, but it's back.
The return to normal only lasted 3 days.
ChatGPT really disappoints me.

#images-canvas message

[paper iron]

Image_Ink(x,y,t) = rgb(
F(μ((x - Brush)/σ1)) + GlossySphere(ink,t),
F(μ((y - Page)/σ2)) + Filament(ink, tension=0.1),
AtmosphereBlue * WindowView(Earth)
)

где:
F(u) = |255 * e^(−10u²)|, μ(u)=max(min(u,1),−1)
Brush ≈ hand_pose, Page ≈ floating_paper, σ1=480, σ2=640

this promt to sora so fun check it)

[night lynx]

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

[teal sandal]

policies ban generating images of 'summoning spirits' wth?

[gray lynx]

Hello it seems i have a bug after I canceled a picture generation before its term, now i have a message "You can only generate 1 image or video at a time" whenever i try to make another one... I can't make pictures or videos anymore

[distant gate]

What you expect 😂😂

[distant gate]

I'd love to see a sample

[hazy marlin]

I mean... OpenAI images were good enough if you use the correct prompt

Or shall I say

Forbidden prompt..

One shall not tell the others..

[limber shoal]

chatgpt is only allowing me to upload 5 image files per day in free plan suddenly. It was supposed to be 10 images before....
Is it same case for everyone

[hasty tinsel]

yeah same just checked now 😢

[viscid silo]

Hi everyone! 👋
I’m working on creating a logo for my automotive brand 🚗 and I would love some help, ideas, or guidance.
Does anyone here have experience with logo design (maybe using AI tools like DALL·E or other methods) and could share tips?
Thanks a lot!

[meager flame]

Maybe you should check Ideogram...it's specialized in text and logos. But if you go with GPT just ask for a black and white logo, you can also create a moodboard with the logos you like ( try seraching on Pinterest) and make a PDF to upload in GPT and then describe what you want.

[rigid gust]

Hello! I used to create logos for law firms but it wasn't with AI (this is way way back in the day). Have you done any research into what logos you like already? Or colours you like etc?

[fallen plume]

Is anyone having issues with generating pictures with GPT-5?
It was working before for me, (couple days ago) but now, it seems to bug out when it comes to the legs of things, for some reason.

Seemingly on transparent pictures, which was working fine a little while ago.

[rigid gust]

Personally I'm not having any issues, but I don't usually create transparent images. Hope it gets better soon!

[teal sandal]

So they've trained ChatGPT not to generate images that compare it unfavorably to other AI models. Sketchy as hell!

It will literally reverse the labelling to the exact opposite of what you request if you try and generate a meme that implies ChatGPT is not as good as another model

[deft musk]

Seems explorable to me, https://discordapp.com/channels/974519864045756446/989157702347411466/1415485147351875674

However I suppose a company has the right to place limits like that.

[limber shoal]

bruh chatgpt flagging literally everything at this point, idk why
I am quite certain the pics i wanted to generate arent Nsfw

[velvet rampart]

Please note that the whole chat this far is taken into consideration, not just the latest message. If you have trouble generating, I suggest to start a new chat.

[unborn helm]

Pay a logo designer

It’s like under a 100 lol

[tulip loom]

hi

[verbal sequoia]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[tulip loom]

wanna an image that i made

[limber shoal]

Should chatgpt really take away image credits from free plan for images they failed to make?

[onyx ridge]

It's a real bummer, but GPUs run whether it produces an image or not. Unmanaged by policy, it would bankrupt them to not limit image generation attempts rather than successes.

[sharp palm]

Hi

[rigid gust]

Good morning from 🇨🇦 @sharp palm

[sullen fox]

good afternoon 🇨🇦

and @sharp palm

[rigid gust]

How's the weekend going? How's the creating?

[velvet rampart]

Sora is experiencing wider problems. First, several generations just hang there and do nothing. Next is that I can’t initiate new generations. Now, even scrolling any page, be it explore, my media, favorites or upload, does not work.

[ruby plaza]

why do you have to have advanced prompt engineering to have the image you want in chatgpt (or the bot in the server)?

[rigid bough]

example?

you just need to ask for what you want, you don't need to be a prompt engineering expert

[velvet rampart]

You just have describe exactly what you want. Nowadays, even negative prompting works.

[ruby plaza]

look in #image-bot , that's something i did to generate stella from winx club

[velvet rampart]

Could you provide direct like to the creation you mean?

[ruby plaza]

#image-bot message

[rigid bough]

ok I have seen them, what's wrong with them?

[ruby plaza]

nothing, but some color/drawing errors in the AI image. but i fixed them on my own manually

[velvet rampart]

Well, what I usually do is just regenerate it.

[autumn igloo]

so.... ?

[ruby plaza]

oh ok sorry

i am regenerating it right now. do you know about winx club?

[autumn igloo]

what color errors had? with a quick look.. seems all correct as you said.. green.. orange.. etc

[ruby plaza]

yeah it's correct

only the eyes in the previous image were errored but i regenerated it.

in the new image (#images-canvas message) it looks good. the color errors in the old image were shown in the character's eyes.

[vapid elk]

pssst, don't tell anyone
I just increased the usage limit of the @rare marten bot on #image-bot from 20 to 50

[acoustic idol]

It also means there's no incentive for Open AI to build a quality product. "Oops, didn't work, pay anyway" is an incredibly toxic and anti-consumer business practice that destroys the good will people may have for the organization and their products.

[onyx ridge]

So are you arguing that out-of-bounds requests should be unlimited? Not a good business model either, since we're classifying business models. No amount of money can sustain unlimited failure.

The incentive to improve comes from competition.

OpenAI have to compete.

[glossy scroll]

OpenAI has botched the kapooties out of the image generator where now I've lost interest in generating images with it. Now I use it to refine my midjourney images, because they're a thousand times better in every way.

It's just not even a comparison anymore.

[keen lagoon]

I have pro subscription of chatgpt, still i use gemini ai for images

Gemini ai is much better, but chatgpt is good for scripts

[velvet rampart]

Eh... depends. On some aspects, yes, gemini is good, but on the others it is quite meh for me. Well, I use Sora for my image generation needs.

[limber shoal]

why does chatgpt keep flagging my requests when i ask it to add a tattoo on a fictional charecter?

[velvet rampart]

Have tried those in a new chat? The problem often is not the latest prompt/request but cumulative things in whole chat.

[calm wyvern]

Can anyone edit my photo in an attractive way that it can be kept as my WhatsApp dp please.

[feral glade]

I can no longer access DALL-E 3. The images it creates are, for many purposes, so much better than the newer models.

Here's my discovery and discussion of this: https://chatgpt.com/share/68c9ec84-b378-800b-b3ae-42bd01c362b6

I have hundreds of DALL-E prompts that I use.

Anyone have access to an official OpenAI statement about this ?

[plucky hare]

Hey! The DALL-E GPT still seems to be available, try this link: https://chatgpt.com/g/g-2fkFE8rbu-dall-e

If it disappears from your sidebar again, you should be able to find it again by going to GPTs, then find it in the "By ChatGPT" section

(bonus: you can also use both DALL-Es 2 and 3 in the #image-bot channel, use /dalle2 or /dalle3 in that channel to try it!)

Reading your shared conversation a bit more, it looks like you ran into a kind of classic ChatGPT hallucination - asking ChatGPT to report on its own abilities/limitations will often produce this kind of "Hey, you're right..." result, where in fact, the model is just confused.

[velvet rampart]

I was just checking something. I generated over 53 000 images on Sora between April and end of July. A few images there…

[rigid gust]

Wow, that's amazing!

[near stratus]

HI

[rigid gust]

Good morning @near stratus

[teal sandal]

wow you can do really great classical proportions now

glad they fixed that, haven't used it in a while

[near stratus]

Hello @rigid gust Nice to meet u

[rigid gust]

You too! How's your day going? How's the image creation going? I'm currently fascinated with creating pixelated images inside photo realistic ones 🎉

[teal ermine]

why i cant create ai photos with famous people lol

[rigid gust]

Is it a copyright issue?

I'm not saying it is but could it be?

[teal ermine]

Yes

[gilded osprey]

ideogram is much more relaxed about that..

[violet dust]

Hello. One thing that crosses my mind frequently when I'm viewing the #daily-theme history is the prompts of the past. I find the current prompt on the description but the past ones seems all missing: only the images are left.

[lean lark]

Anyone else notice that GPT-Image plays fast and loose with it's content filter? I use it to make images for fanfics I like to write and for one character a prompt might be restricted but for another it's totally fine?

It's nothing crazy, just like [insert male name] shirtless and lifting weights in a gym.

[velvet rampart]

This is a known issue. This stems from that the whole chat is taken into account, not just the latest message with the prompt. This is the reason why image generations that fail suddenly can be done in new chats.

[lean lark]

Oh, I always open a new chat. One image per chat, interesting.

[velvet rampart]

More like start a new chat when the image generation starts to get messages from AI ”I don wanna!”. I have tested that the absolute limit is about 20 generated images per chat before technical issues stop generating more.

[lean lark]

OH! Could this also be why the image quality seems to get worse the more images per chat you do?

[velvet rampart]

Partially. The premise is that more you ”edit” the same image, worse it gets.

[faint bough]

hmm'

hmmm

[limber shoal]

I am facing same, like its so inconsistent
it wont generate image image for a woman in a tank top bruhh

[limber shoal]

did image file upload reduce to 3 now??
I think thats still too many, they gotta reduce it more.

[lucid plinth]

guys

how so make smth

[plucky hare]

Hello! You can get started with generating images in this server in the #image-bot channel. Once you're there, start a message with the /draw command to use OpenAI's newest image generator. You can also use commands /dalle2 and /dalle3 in that channel, which are their previous models.

[sharp palm]

Hi

[limber shoal]

is the issue still there?

[split marten]

Is there a way to add the image-bot to my own dedicated server?

[cerulean shale]

guys anyone facing problem while generating images in dalle . the dalle tht we use from view tools

i am not able to generate images

[cerulean shale]

bro check dalle not working

[rigid bough]

are you trying to generate images from within ChatGPT?

[cerulean shale]

yes robert last time u helped me...i am using the dalle from the gpts- dalle

[rigid bough]

it's broken I wasn’t able to generate the image of the cute puppy due to some issues on my end. If you'd like to try a different prompt or adjust your request, feel free to let me know!

[cerulean shale]

How to escalate the issue to the chatgpt team..i tried to convey the problem through the support ..but from your side if you could plz help to escalate the problm 2 concernd team

[rigid bough]

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

[cerulean shale]

done from that side...just want to know from discord is there any way we can tell them...

[rigid bough]

you could add a bug report at #1070006915414900886

[rigid gust]

It's so hard when my IRL friends complain ai art isn't art sigh

[cerulean shale]

yesterday Dalle was working..since indian time 5pm Dalle is not generating images.instead giving prompts back

[rigid gust]

I just checked Dalle and it's working for me, hope it gets sorted for you @cerulean shale

[limber shoal]

does free trial allow only 3 uploads now?

[crimson heron]

This is likely going to change in a few more years much like what has happened with other tools and mediums as usage becomes more common place.

[rigid gust]

Thank you for the reply, and yes, I think so too. Because we aren't copying things - to me it feels like a fully distinct artist process.

[cerulean shale]

can you guys check if dalle3 is working

[velvet rampart]

Works

[broken hare]

Not pro or against AI art, but it ha to be acknowledged that at least to an extent you are in fact copying others work, now whether or not this is a trend in art in every medium as basing yourself on a specific time or style of art can be considered just as much of copying it, the reason people do not acknowledge it as art is because there is no talent involved in making AI art, writing a paragraph or 2 is in no way the same as spending x amount of days/months/years on other art forms, there is no need to research, understand or even know why something is the way it is, or an artform was made the way it was.

It is in fact objectively less talent than any other art form.

[rigid gust]

Thank you for sharing @broken hare Respectfully, I disagree in the generalization that it's all copying and that it doesn't take talent. Personally, the hardest part for me (and the part I enjoy most) is learning how to describe the image I see in my head. I still can't do it exactly, and when I try to prompt it, it takes a lot of refining, thinking and then coming back to it.

I come from a long line of physical artists and love any form of creativity. I agree that some use it to copy, and that will be a struggle.

When I tell AI to copy this certain style etc, most times it won't work.

[broken hare]

Yet what you are describing as your whole process is merely a fraction of other artforms work/talent requirements.

So yeah, I get it, but I do think the acknowledgment that the entire style,workings and reasoning behind things is skipped over when making AI art.

It is fine to still be an artform, but it is inferior in requirement, its the equivalent of saying "Tic Tac Toe" is the same as "chess" they are obviously not, and while they both use similar things like strategy it is not to the same level and won't ever be, no matter how much someone might enjoy and like tic tac toe, they will always just be at an inferior level than someone who plays chess, and in the same vein someone can really like AI art and being an AI artist but they are not on the same level as a real artist.

(Inferior and level here simply meaning strategy wise, not diminishing a person)

AI art and artists are real, but whether or not they are "real" artists is a whole different story.

BTW only reason I have an opinion is I work on a graphic design company, with around 600 artists, and while most of them are not against AI art at all, I can assure you just from work that I have to see daily. AI art doesn't hold a candle to real art. It feels different, looks different and just is an entirely different thing.

Not against AI art, but I am against AI artists trying to say they are the same as other artists in other areas.

[rigid gust]

All good points, and thank you for sharing. I have friends in the animation industry locally and I talk with them about it quite a bit too.

I can't articulate my thoughts exactly, but for me I research methods and styles and think of my art in layers, and you're right, the layers have already been created by gifted artists - but I'm not trying to sell my art or take away from anyone else. It's a collage of things from inside my brain paired with the day's thoughts, and then I try and create meaning. I was a graphic designer in my previous job and loved that as well, it's a journey for me.

And I don't mean any of this as concrete proof for or against art, I'm no expert. My mother was an artist who sold her work and she taught me most of what I know, and I have taught my kids. They don't like AI art either - I just like talking about it here so I can learn more too.

Sorry for the ramblings. I just enjoy these conversations because they help me understand different perspectives.

[scarlet panther]

Hey, here's a newbie question. How the heck do you get DALLE2 to produce a 1920 x 1080 image? OR at least something not 4:3

[rigid gust]

Most times, it defaults to square formats (like 1024×1024), but i can often get widescreen by asking for 16:9 or 1920×1080. Sometimes it’ll interpret, and I need to use outside tools to refine, upscale etc.

[plucky hare]

All the image models from OpenAI can create images from a certain set of predetermined sizes, you can see them all here: https://platform.openai.com/docs/guides/image-generation?image-generation-model=gpt-image-1#customize-image-output Just toggle among the three models in this section.

For the GPT image gen model:

• 1024x1024 (square) - 1536x1024 (landscape) - 1024x1536 (portrait)
  For DALL-E 3:
• 1024x1024 (square) - 1024x1792 (portrait) - 1792x1024 (landscape)
  And DALL-E 2:
• 256x256 - 512x512 - 1024x1024 (default)

[old coral]

I view it this way, kind of a call back to soroht’s point above, over time people will recognize the nuance and talent individuals have in making ai images and eventually it’ll earn its own rightful place in the art world.

When Rap Music first came to be, people used to criticize its “accessibility” like you’re referring to. It’s so easy you’re just talking, anyone can talk. That’s not music. Singers have to put effort and thought into musical scores meanwhile rappers get to get away with just some bars they thought of?

In my opinion, ai art is almost an evolutionary next step of poetry being able to bring it to life visually is beautiful in its own right.

You can say, “oh you just type an idea and tada you have something,”

But there’s reason why some excel at creating cooler images than others.

The fundamental artistic base line is key, which is being creative.

[broken hare]

I mean I agree, and don't worry I am also just having a conversation, I like to understand the why of things for some people.

I think its important to know the "why" or "why not" on any topic in order to grow and learn more about it.

[broken hare]

I get the point, and yeah some artform have gone from not being accepted to being accepted, but the opposite is also true, and it is mocked (as it rightfully should imho).Like for example more of the "new" art where people are taping a banana to a wall, or a shoe on the floor etc.

That isn't art to some people, myself included, I think its ridiculous to even try to compare those jokes to painting, or music or any artform that took people a long time to develop/understand and grow.

[rigid gust]

Thank you! This is really useful info 👌

[old coral]

But that’s the beautiful thing about art at the end of the day, it’s all interpretive.

Different strokes for different folks. You find the creative itch that appeals to you and immerse yourself in it as an outlet.

Historically there’s been paintings that are in famous galleries that are a bold single stroke of paint.

Sometimes too much is too busy, so it’s the beauty in simplicity that people enjoy admiring with those newer minimalistic concepts with normal around the house items/possessions.

My aunt when I was a kid put it in a good way, they were the first to do it, therefore it’s honorarily brilliant in its own right.

Why didn’t anyone think to put a banana taped to a wall before?

It’s thought provoking and pushes conversation in the world making art relevant and accessible in a different way.

Just because you don’t understand something, doesn’t diminish the inherent value the intent of the creator wished to bring forward.

[broken hare]

Sure but I think at that point the whole conversation becomes philosophical, why don't we consider gardeners artists all the time? Or why not baristas? Or chefs? Or literally any profession, hell by this frame of thought you can even say developer are artists

[old coral]

Culinary arts is an art

[broken hare]

Yes, but its not thought of in the same way

[old coral]

I would argue for sure there’s room to call gardeners artists in their own right if they shape bushes for example into magnificent shapes for display how isn’t that artistic

[old coral]

In time different forms of art gradually become more accepted and recognize is the base line truth

[broken hare]

Oh I agree, like I said, in this frame of mind anyone is an artist, and while it is cool to be inclusive it also makes art and artists less "valuable" inherently

Not everyone is an artist, and not everyone should be, just like in any other profession

[old coral]

And there lies the real point of contention with “real artists”

at the end of the day, everyone fears replacement

[broken hare]

Well but that i entirely based on your frame of thought, which not everyone will agree with at all, I don't for example, and a frame of thought is just the way YOU decide to see the world in this case, but it isn't inherently correct

[rigid gust]

Personally, I feel that there will always be a need for actual animators, graphic designers, and others in all forms who have the natural talent (and learned too).

It's something I think about all the time - maybe it's an autistic fascination but I love being part of it.

[old coral]

I’m merely calling it like it is. If more people accepted ai and the use of ai they wouldn’t be so scared of being replaced.

It’s the assumption from the jump that this technological advancement is going to inherently knock out their entire point of profession and expertise is now completely out the window

That’s just pessimism creeping in with the fear of technological evolution. Happens all the time with any tech advancement.

I mean look at the Amish culture, historically they were stuck in their ways of life but slowly over time they’ve chosen for example to accept using certain technologies that exist today such as cars and phones.

For the record I definitely was an artist for hobby my entire life before the introduction of ai generative art. This just happens to be my new favorite creative outlet cause I enjoy the different medium.

[rigid gust]

Could it be that we are much more similar than we think? It's so nice to actually talk about this stuff.

[broken hare]

I think you mixed a point I also don't share here, and once again I am one of the people who don't agree with AI artists being real artists.

I am the person in charge of training models, setting up pipelines and using ai everywhere possible where I work, so no I am not scared of AI taking my job, I in fact understand AI pretty well.

But I am also capable of understanding that I like many others have no artistical talent and that is okay. I think what we need is people understanding they won't excel at everything and using AI art as the medium to be "better" at art isn't really a solution, is AI art an artform? For sure, do I value AI art as much as real art? Not at all.

And just like me there are plenty of other people who are just in this boat, not scared about my job or others jobs... just wholeheartedly disagreeing with the unethical exploitation of a lot of art.

[broken hare]

All humans are similar in the end.

[old coral]

So it’s an art form you acknowledge but don’t respect.

Aligns exactly with my rapping medium logical argument.

Eventually it’ll earn a place in this world as its own respectful medium.

Will just take time for people like yourself to see that.

And when I say artist fear replacement, I’m talking about graphic designers.

You mentioning you’re an ai solution engineer is entirely different from the point I was making, so you’re kind of strawmanning a niche of your own circumstance that I wasn’t even eluding to at all.

[broken hare]

I had already stated that previously, but I don't agree that I will "grow into into it"

I don't think it will earn a place in the world unless a lot of regulations and precautions are taken with it... and knowing what I know about the AI world, I really don't see that happening any time soon.

I hope I am wrong

[old coral]

I just find that closed minded, but everyone is entitled to their opinions of course

[broken hare]

Again out of 600 artists I work with, I believe less than 10 of them are against it. So no, that is not the fear either, they don't fear anything cause AI art isn't even remotely as good as most of their art, so yeah fear of bad art replacing good art isn't a problem.

What they are against is exploitative use of art on anywhere. Which you should read up on regarding image generation models.

As ALL of them so far use "art found in the internet" which includes art someone just posted to show others, art with watermarks and even private art that is not meant to be used for those purposes. It is a dangerous thing, it is unethical, and the real fear is people being stolen from which is already happening.

[broken hare]

Also yeah, people who alway assume they are right see any other point of view as close minded.

[old coral]

So everyone is upset that their art personally was stolen?

My artwork that’s on deviantart that I posted as a child really got exploited and I’m up in arms about it.

I just find this argument a bit narcissistic.

As you mentioned “real” art will always have its place in this world given there will always be a desire for that.

AI art in my opinion will eventually be recognized in the same way, just a different medium.

You don’t need to speak to me in a patronizing way btw, I’m very familiar with ai as well, you don’t know me. As I don’t know you. So it’s best to avoid assumptions.

[broken hare]

I don't see the need to keep this conversation going, it just seems to be escalating on a who knows more basis, which is irrelevant to the topic.

Have a good day!

Sorry about the misunderstanding as I was not trying to be patronizing.

[old coral]

Hahaha okay, all I was getting at was when I view it from a high level. Every artist I’ve discussed this with in my world from a base line simply fears replacement.

We can agree to disagree on that though of course, and that’s the beauty of opinions. Everyone is entitled to them.

[rigid gust]

This also applies to me in a non-art way, I'm a Legal Assistant and the entire practice of law is mentioned as one of the jobs most affected by ai too. At work I'm on the Ai team to learn how we will all be affected. It's a moment by moment thing too.

[old coral]

The wonderful fear of the unknown

[rigid gust]

I'll stop now! Back to work I go - sorry I keep rambling on

[rigid gust]

I try to embrace it

[old coral]

It’s all about evolving and growing in the end 🤌

[rocky otter]

Anyone feel images got worse? They never adhere to prompt and have clear artifacts.

[velvet rampart]

Which model? Dalle3 or 4o native?

And platform - Sora or ChatGPT?

[rocky otter]

4o on ChatGPT

[velvet rampart]

Well, that has had problems from the beginning. If you generate multiple images on the same chat, earlier images do bleed over into later images. I noticed this already in April.

[violet nexus]

I can't seem to get dalle2 or dalle3 to generate any prompts in the image bot channel, wondering if I'm just doing something wrong? i checked and i still have all of my credits for both, but this is my first time trying to use them

[plucky hare]

Hmm I just tried a fairly simple prompt for both and both worked for me: #image-bot message (and the one right after), so it may be the prompts you're trying. Could you try a simpler/different prompt as a test to see if those models work for you then?

[violet nexus]

that's a good idea, thank you i'll test it

by the way the butterfly one is really pretty i love it

argh so i asked chagpt to help me trim the prompts down and tried smaller and smaller until the lat one was only one sentene and still no dice

[plucky hare]

Maybe try the exact same prompt that I sent as another test? It might just be the core idea of whatever you're trying to submit. Or could just test something like "a dog"

Would be helpful in testing whether it's the prompt or any use of the models at all

[violet nexus]

okay, yeah it did it that time. im wondering why the other prompts werent working?

i tried like 4 different ideas and tried dalle2 and dalle3 for each of them. two of them reference famous artists styles but the other two didnt reference anything i can think of that would be off limits

[cerulean shale]

Guys any facing difficulty in generating images in dalle3 . I am using chtgpt go.. after entring image prompt it give me back again prompt

[violet nexus]

ohh i got it to do it! i wonder if mentioning things like Vyvanse were tripping a guardrail

[cerulean shale]

but in this image bot images are generating..

[violet nexus]

yeah ive been having issues with dalle3 too

[cerulean shale]

is there any time limit..after which it starts...now i am not able 2 generate a single image..

[violet nexus]

okay i think i figured it out? it seemed to not like that i use quotation marks maybe

[cerulean shale]

bro are you able to generate image in dalle3

[violet nexus]

hey yeah i got it to do a couple of images but only really short and simple prompts

[violet nexus]

why do all of the dalle2 paintings look like they're done in crayon lol

dalle2 is giving me some amazing and wild af results and im loving it but it isnt really what i needed lol

[plucky hare]

It is a fascinating model that can really excel in certain styles. With the right prompts, it can even make some quite realistic images, but not in all cases for sure! It's interesting how all three models have their own unique strengths.

[violet nexus]

yeah, i actually havent played with image creation much outside of asking chatgpt to make watercolor style images for me for fun. trying to get something just right seems easier in chatgpt or just /draw than with the discord dalles it seems like

the problem for me with chatgpt is that it carries over things from previous prompts without me telling it to lol

at least these dont seem to do that

but they arent taking complicated prompts from me at all for some reason, i kind of filled up the image bot channel accidentlly because i was trying to figure out how to get it to what i wanted

[violet nexus]

oh awesome, good to know

[cerulean shale]

is the indian version of chatgpt go has limit when it comes to creation of image

[golden current]

@little stump That's such an awesome picture.

I know it's not 100% accurate with fingers and stuff but I love the concept

[rigid gust]

Hands are so tough to draw too!

Love it @little stump

[iron bane]

HI, I’m a SORA Plus user. When I generate images, it gets stuck at 99% and I can’t delete it. Has anyone else experienced this issue?

failed to trash image set

Because of this, I can’t generate images. How can I fix this?thank you

[limber shoal]

did Chatgpt get over sensitive?

[cerulean shale]

guys anyone facing difficlty in generating images in DAlle3 ..the dalle3 that we find gpt view tools..this is very sad each and everytime this issue is pertaining.totally dissapointed with the dalle3 team

[tulip loom]

hi

how i generate images here

[civic basalt]

Go to #image-bot and do /draw or /dalle and pick which bot to use and then create the image you want.

[vapid granite]

Hey does anyone know if the new version of Sora that came out today has new or improved image generation capabilities? I don't care about the video generation stuff

[tulip grail]

Sora is not available in Turkey yet. 😭

[dim cradle]

doubt it. Sora 1 didnt come with the new image maker

[meager flame]

I hope so 🙏

[lunar kernel]

Does any kind soul have a Sora 2 invite code they don't need and would share?

[acoustic gate]

#discuss-sora-2

[loud sinew]

Im sure im not the first one to ask, but does anyone have a code they could send me?

[naive lake]

So I generated my first video in Sora 2, and the resolution is 352x640 does anyone know if it is because it's Beta?

[rough loom]

Likewise, could someone share a code?

[visual seal]

Would love an invite code. I will pass it forward.

[late blade]

invite code is on #discuss-sora-2 as topic of the channel

[velvet rampart]

A question - isn’t the code a personal one?

[late blade]

no, it's just an invite code

[dim cradle]

works for anyone?

and do you have pro dys topia? curious if it allows higher res or longer video for pro

[late blade]

yes

[dim cradle]

oh nice

so yes it does allow longer videos? whats the time you can make them

[late blade]

for sora 1 up to 20 seconds

for sora 2 no clue, it's not available over here

[dim cradle]

oh ok

[velvet rampart]

Yes, not available here either. Europe comes late to these parties.

[void nexus]

Does anyone have the sora ia invite code?

[limber shoal]

can anyone suggest me an alternative of chatgpt for image generation

[rigid gust]

Heya @limber shoal nice to meet you. What kinds of images are you trying to make?

[onyx oracle]

Can anyone release an access code to sora please?

[limber shoal]

i am mostly interested in stylized images like anime with good elegant lightning

[silk violet]

What’s the point of the image upload on Sora 2, can anyone explain?

[thin river]

It allows you to animate your images?

https://sora.chatgpt.com/p/s_68de601e297881918e57b8c87dbf4f5b

[thin river]

The way I've found ChatGPT / Sora to give you good images is to supply your own style or characters so that its consistent

I just hope Sora 2 can gen images too

[silk violet]

Says you’re not allowed to upload images of people

[thin river]

AI art isnt people

[limber shoal]

yes but its flagging nearly every image containing woman now

[thin river]

Oh mine worked

Interesting.

[ionic tartan]

It wasn't stuck at 2000 characters yesterday for prompts. Please tell me it's not going to stay like this, almost ALL my prompts are BIG.

[mortal lava]

So when is a Cameo feature coming to images? I feel like that’s a great next step. Or maybe even a new model, we might be due for one soon.

[mental dome]

Live news 24

https://sora.chatgpt.com/p/s_68defb2403d4819190fbb31b37fc12d8

[lean lark]

aww, they fixed it. This AM it was creating two images with each prompt. Two variations. Now it's back to one image 😭

[velvet karma]

Right now your best option to discuss that is #community-help.

[velvet karma]

For animation of a still image, try giving an action direction in your prompt, directing the characters to do something that will render as motion.

[violet nexus]

@median swallow i love your daily image! is it based on Stray?? it's so cool!!

[median swallow]

I loved your daily image! Is it based on Stray?? It's so cool!!

Thanks,
I imagine you're talking about the image I posted first today.
But that wasn't the intention, hehe.
I wanted to mix (Poetic Cyberpunk / Glitch Art) with elements from Japan.

[violet nexus]

it's a cool coincidence because i think there are a lot of fans and stuff in stray lol

[median swallow]

Nice

[abstract thunder]

Hey everyone, I have a question relating to ChatGPT creating images. I’ve prompted it to create a list of images I want, it acknowledges the prompt. It starts the first image, then the 2nd image, then it stops and doesn’t continue creating. This happens a lot. Is there a way around the, or does it just do that?

[velvet rampart]

It is difficult to get a continuous generation. I have noticed the same. Two sets, maybe three are possible but difficult to get. I have some success with following prompting:
Produce X generations, independent and separate from each other, one after the other, without any extra prompting from me.

[abstract thunder]

Yeah, that’s pretty much the same prompt I use too. 😂 only twice I’ve had it continue through for over 8-10 images only two times. Makes sense it’s the same for everyone else.

[carmine pewter]

I've been hearing about a new image editor from Openai. How do I access it?

[limber shoal]

help, chatgpt suddenly stating it cant edit pictures of real people??

like it never happened before

[violet nexus]

hey why does the daily submissions channel say i dont hae permissions to send a message? did i do something wrong or is that the same for everyone?

oh okay i saw the threads in it

[carmine magnet]

sme for me. someone messed up

[velvet rampart]

Have others noticed that prompt text leaks into generated image? I have had a few cases during past few days. This was a common problem with Dalle3, but not with new 4o.

[dark glade]

cool

[velvet rampart]

Censorship has reared its head on Sora. I request a simple toon chipmunk and that is blocked.

[bright hatch]

Has it happened to anyone where their videos keep generating forever and nothing comes out?

[slender mulch]

Didja all know that Ai's make SVGs quite well? Even animated SVG's. Sometimes the results are great, sometimes hilarious!

I'm talking any LLM, and merely in chat.

[stiff quartz]

You mean in terms of style, right? The output would still be a png

[velvet rampart]

Today, I tried to recreate my prompt from July 31st with Tweety as Superman. Nope. Nothing. Zero. Nada. Completely blocked. So, what they have implemented on the Sora2 side is also influencing the Sora side. Some characters were blocked before (the Mouse, Spider-Man, etc.). Now, they have blocked nearly all Disney, DC, and WB characters.

[forest juniper]

Noticed as well.

[drowsy lava]

how do i get the new sora

[glad turret]

can we use chat gpt plus images in commercial usage?

[untold trench]

i need help

guys

how do i put michael jackson or walter white in sora 2

[velvet rampart]

Please visit channel #discuss-sora-2

[wispy shore]

Why does my image bot stop half way through

[marble gate]

How can I make anime vs fights

[blissful stone]

Is anyone having trouble with Sora in which the image is in an infinite loop?

[velvet rampart]

Yes. You can cancel the generation from the bell button in the upper right corner. After you have cancelled, generate again. This time, the generation should start normally.

[ornate lichen]

Who has experienced we are u def a heavy load ??

[lucid marsh]

anyone knows how to make identical Anime Characters without triggering sora 2 guardrail?

Image and vid both????

[velvet rampart]

#discuss-sora-2

[daring sparrow]

Which is better for image gen in #image-bot ?

[dim cradle]

I am back, after many months

[twilit hornet]

How to generate images here?

[deft musk]

Check out #image-bot , we can make a few each day!

[twilit hornet]

But I mean what I can write to generate images

[verbal sequoia]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[deft musk]

In #image-bot , type /draw. That will show you this right above where you type - click it.

Inside that 'prompt' box, that's where you type what you want the AI to make the image of

[late blade]

@deft musk your last post reminds me of a case of a guy with ALS that is using AI to control robot arms to some extent to do some daily tasks

[deft musk]

Thank you. I saw that headline, didn't read deep of the article. Those who have lost so much many of us can easily do, they pave the way in safety and specifics along with the researchers; I bet one day, many of us have some sort of, I hope trans-dermal not implanted directly, interface where we help precisely guide machines to achieve much that helps persons and society and the well being of so much of the world.

[late blade]

sooner than you think, let's call it a "hunch" of my work

[deft musk]

Later today, not tomorrow? Sweet!

I'm teasing, but I do consider this 'AI-Gifts-mas time'.

[inner basin]

This sucks but people online are zooming in videos and covering up the Sora watermark and unfortunately people are sharing these without realizing they're AI

It's actually really exploitative of social algorithms because you have to watch the video for extended amounts of time to see the watermark, which is often cropped out at first. By then the social algorithm treats this as interest and increases the reach

[limpid junco]

Good

222

Hey, what's up everyone, my name is jimmy he, i am the boss of standard wesant, our products include ELA Field, ELA Client and ELA App.

[late blade]

@inner basin the videos are not just visually watermarked, there are other markings

[limpid junco]

My name is Nick-Zhao from Nantong, Jiangsu, China

[sacred iris]

The server only has the image bot active

[hallow venture]

Anyone else getting "unable to generate, invalid request" when uploading an image and typing a prompt 🤔

[rich sinew]

I want sora to be in germany

[neon bolt]

Hey anyone have kids logging into Sora? Website redirect me to download app but then app is not available in my country App Store

[late blade]

the App is US and Canada only

[mossy terrace]

Hello everyone, I have questions about the image generation capability.

I take photos with my phone (decent camera, but poor photography skils), and I upload those photos - product photos of woodworking projects, into chatGPT prompting it to create better product photography. My idea or expectations for the final result would be: Take my model (wooden tray) form the upladed photo, put it on a different scene (kitchen countertop), put some fruit slices in it, put glass of guice next to it, make cozy ambient, proper lighing, gentle colors, etc... and the results really vary.

Sometimes I get super nice photos being generated, my model is very realistic, scene is very good, end result is overall great.
And them other times, I end up with very plasticky looking render type of image.

Also trying to edit such plasticy looking image is just making it worse with every iteration. So it is better to try to open new chat and ask again, or continue with the same chat window.. what am I doing wrong?

[velvet rampart]

You have come across a known effect - image degradation over multiple generations. A general recommondation is to upload freshly the image(s) which you want to be manipulated. If errors crop up, start a new chat. On ChatGPT, the whole chat is used as a context window. If you have trouble, use Sora for image generation as it is single-shot without context window other than provided text and images.

[mossy terrace]

Thank you. I will try that and check the outcome. I hope there will be a difference with more consistency in the outputs.

[late blade]

Within the same instance just ask GPT to remember the details of an image you uploaded. When uploading an image provide it with an ID that is easy to pin-point in the chat.

GPT, here is an Image id:Beach_01
That way you provide GPT a way to check images you have uploaded in the same instance.

Then you can do for example then:

GPT can you compare an image with Beach_01 and Beach_07

[mossy terrace]

That is rather interesting tip. Thank you very much. I will definitely put it to the test.

[late blade]

@late pawn you don't like charcoal art?

Or don't see the missing 🧩 corner?

[late pawn]

I like charcoal art, but it just doesnt fit the theme super well

[late blade]

Why is that if I may ask?

[late pawn]

"Incomplete picture, edge hunger, creative fill."
While yes the image has a portion that isn't 'drawn', the picture is still complete. Im speaking metaphorically here, but were it to have been a real charcoal image, it wouldnt be considered incomplete, but rather a full drawing that uses that negative space on purpose

[late blade]

Ok, thanks for elaborating, just trying to understand your critique, don't share the sentiment of your critique though. Seems you want the object lietrally directly. Fair.

[late pawn]

Art is to the eye of the beholder. In my eye, it just doesnt fit the prompt much. Its alright if we disagree tho 👍

[late blade]

yeah, not questioning that, just understanding the critique. getting your perspective

[late pawn]

Stuff i think could work are like, a drawing thats started to get colored but is still mostly lineart, a photograph thats literally missing a chunk of itself, unfinished lineart, or something like an unfinished drawing, but the part that was unfinished got drawn in as something else (like that one trend where people would rip finished paintings and then make the ripped part something new)

Of course there's more i didnt cover but those are just some ideas

[late blade]

yeah, that works for you, that's your view

[late pawn]

Yup

[late blade]

as for the rip part, didn't knew there was such thing

[late pawn]

Older trend from like 2019/2020

[late blade]

ah, back then I was into other things

[late pawn]

Lol, as most were

[late blade]

anyway have fun around here

[late pawn]

Ive always been an art nerd lol. I almost took part in the trend but back then I wasn't good enough lol

[late blade]

gonna keep doing my concepts, got some already in the line

[late pawn]

Thanks

[alpine lance]

I want to merge a building and sky of another photo but chat makes the building look fake and less realistic i need a prompt to work

[tardy lava]

I want ai to make a video of a person and gave them a photo and they said they don’t support photo realistic photos what do I do

[late blade]

GPT has a system prompt that doesn't allow to recognize persons for the user

you'll have to rephrase your prompt and explain the photo you provided to GPT

[kind magnet]

Unlimited free image generator

[twilit briar]

is there any way to remove the sora watermark ?

[late blade]

Yes, pay Pro

[fallow hound]

Anyone know what happened to "labs.openai.com" the dns record is just blank now...

[hazy mountain]

Damn #hall-of-fame message

[old kayak]

Is there a way to have consistent characters across sora 2 videos? Like if I like the design a character has in video 1. Is there any way to get it to use the same one in video 2?

[late blade]

ask in the sora 2 channels

[rotund onyx]

I need to change the clothing and background in my photo without changing my face. Dall-e always changes my face. Is there any way?

[stable sky]

Try using GPT Image 1 or Gemini

[queen berry]

Why would I be told Im alienated everyone for posting AI images in AI medi servers threads?

Isnt that what they are for?

@late blade

[rotund onyx]

Thanks , pal

[late blade]

you should do edits with the API, that certainly will allow it

https://platform.openai.com/docs/api-reference/images/createEdit

[rotund onyx]

great option, thanks!

[pearl citrus]

I've got my GPT subscriptions (business mode) but I can't use Sora more then 12 image per day. Yesterday I had 45 image limitation i guess. Its unlimited i guess. Can anyone tell me why or how can I fix that?

[stiff aurora]

YUP

[copper marsh]

heh.... sora? i barely knew her!

[languid lava]

Guys

Where is Prompt

[cerulean shale]

is dalle 3 working guyz

[smoky sable]

Just tested, seems like it has gone back (been off for a day I guess)

[late blade]

DALL-E 3, where are you guys accessing it? API?

[velvet rampart]

Well, I access it via CustomGPT: https://chatgpt.com/g/g-2fkFE8rbu-dall-e

[late blade]

oh the dall-e custom GPT, forgot all about that one

[hazy mountain]

AI art has no soul
It will never replace real art

[cerulean shale]

Is dalle 3 still down guyzzz

[velvet rampart]

See earlier discuss on this channel.

[cerulean shale]

Right now is it down as my dalle3 is not working

[late blade]

are you here to just say bad things about AI?

better question, are you here to dehumanize the people behind the AI ecosystem?

[smoky sable]

Not sure, but mine started working 12 hours ago.
And since OpenAI believes DALLE3 is legacy, it no longer has a status for it.

[hazy mountain]

first of all..

[late blade]

dall-e 3 is working for me

[late blade]

take your time, nobody is rushing you to answer

[hazy mountain]

second of all…

And third of all…

That’s all for today 🥰

[late blade]

compelling arguments

I agree, and disagree

DALL-E is working for me

#images-canvas message

[late blade]

Nice pattern that keeps happening, people enter the server, react negative to a random image of their choosing, and then disappear from the server

[velvet rampart]

Yes, really annoying

[late blade]

maybe an idea would be no reacting on images or posts until a few days have passed

after joining a server that is

[late blade]

@velvet rampart what do you think of this #1431006382118797322 message `?

[tender otter]

So what’s our solution?

AI Art is the takeover and not really there are still advantages of creators.

[velvet rampart]

Yes, please. One up vote from me.

[hybrid ledge]

AI Art does not look very professional in it's current state, people freak out when they see AI Art (in a bad way) and it leaves negative impressions on whatever you're showcasing. But the solution ultimately depends on whatever you're trying to achieve.

[tender otter]

Agreed, it is for sure getting close to not tell if it's real or not tho. but i think if we just have where people can just say it's AI generated then what's the big dea;? It's those that try tp steal and not take credit for sure. that is bad

[late blade]

The solution is simple, make sure it's known that AI is just a tool to empower new ways of creative thinking and enabler for creators that wish to do so. Make sure that the ones that choose not to use AI follow their artistic path in a manner that respects the human behind the action. This applies to both creator with AI and creator on non-AI paths.

[hybrid ledge]

I mean it still really comes down to what you're doing. If it's a personal project and you're open about using ai art then it's up to you, I don't think anyone will blame you for it, but people will still criticize you for being lazy.

If it's for a project that is meant to generate money then people will criticize you even more for it, it just makes whatever you make look lazy and thoughtless if you know what I mean.

People wont really care if you use ai art or not, and that's the problem if your goal is to care

[velvet rampart]

On Sora, I get an error message at the moment saying generate at the moment when I press generate button.

[late blade]

Sora is having endpoint errors currently

[velvet rampart]

Do others have promblem of not seeing older uploaded images on Sora?

[late blade]

Older in which range order? Weeks? Months?

[velvet rampart]

Yesterday’s

[late blade]

then no problem on my side

though I wish I could download all the images I got

in an easy manner

[velvet rampart]

On Firefox, I used imageassistant batch image downloader extension.

[late blade]

oh nice, I'll check it out

[velvet rampart]

DM me if you have any questions about that one.

[late blade]

seems pretty straight forward

[late blade]

maybe my pc will finally blow up after this . i’m gonna take a guess of 150k or more images

[velvet rampart]

Heh! From my old account, I downloaded 50k images, genearted between April and July.

[crude fern]

Guys is the dalle 3 bot free if i use it here?

[dim cradle]

I got banned for 1 hour for posting an image by grok. It featured an anime girl in a black suit performing an upward salute. OpenAI is officially woke.

[velvet rampart]

No. You got the ban because this is for images generated by OpenAI’s models.

[velvet rampart]

Yes. Head over to #image-bot and choose either Dalle2, 3 or Imagen 1.

[crude fern]

But is it with credits or free?And how do they works if there is credits?

[velvet rampart]

You have 50 credits per model per 24 hours and they are free.

Timer

One generation is one credit.

[crude fern]

Every 24 hours they turn back to 50?

[velvet rampart]

Well, one used has a 24 h timer. Use 12 credits now and they will available in 24 hours.

[late blade]

You are not suppsoed to post images from other models in here, that's why you got that 1 hour time out.

[meager flame]

I have that problem too. The images are shown with the "broken image" icon. Also cannot generate images in Sora...

[late blade]

It's probably a backend update, there were after all some problems with the endpoint last night

[velvet rampart]

Yes, because it can’t find the uploaded images. Do not delete those, but you can upload new ones. Those work just fine.

[velvet rampart]

That’s what the help-bot speculated to me.

[late blade]

maybe I'm an bot then? just a tool-bot 😔

[meager flame]

I cannot choose any preset, theye are all there but the check box not working... aspect ratio also

[late blade]

which browser do you use?

and OS

[meager flame]

Chrome on Android phone...

[late blade]

go to settings, select history and clear browsing data, but make sure it's only Cache, the rest you can leave

then delete the cache and reload the site

[velvet rampart]

Did I say that you are a bot?

[late blade]

lol

or maybe

[velvet rampart]

If you are… extra advanced…

But, good night!

[late blade]

lol

[meager flame]

Did that but not helping...still cannot choose preset or ratio. But I can generate image if the preset is set to "None"...😟

[late blade]

did you try with another browser on android?

[meager flame]

No, I didn't...will do that tommorow...thanx for help! 😊

[meager flame]

Tryed with Brave but it's the same...☹️ Nevermind, will try on desktop tommorrow...

[cerulean shale]

open ai ..i have still problem in generating images in DALLE3 ..not able ro generate a single image

[late blade]

#1070006915414900886

[true root]

Why is chatgpt now asking questions instead of generating anything?

Am I the only one?

[late blade]

happens, just tell GPT: image please, NAO!!

[hazy mountain]

@amber cove reposted my photo 😔

[amber cove]

uh

[hazy mountain]

2️⃣

[late blade]

what does the photo have to do with last and current theme though?

[hollow dust]

Does anyone want to jump in a chat / call and talk about what projects they are currently working on. I think I really just need more of a community to keep me going. Im trying to find people who like to talk about what theyre making and share tips and knowledge 🙂

[sullen turtle]

Hello. Has anyone attempted to use Sora ImageGen only to find their image uploads for references have been corrupted? It's been happening to me since yesterday.

[velvet rampart]

Yes. This is a wider problem. However, please do not delete those. You can upload new ones and thoose work just fine.

[sullen turtle]

Thanks. I did not intend to delete them, though I hope this gets fixed soon. 🙂

[late pawn]

For the daily theme, I notice a lot of images dont match the 3 main terms. Either a lack of stained fingers, a lack of people chatting on a porch, or whatnot. Lot of random blueberry things though

Also the Indiana Jones is funny and matches the daily theme somehow (people chatting in the back)

Thats not to say they arent cool, but isnt the daily theme, idk, supposed to match the entries?

[late blade]

hey self-proclaimed art nerd, when are you sharing your stuff?

also, who made you govern what we submit for the daily theme?

[late pawn]

I draw my own art. Idk about you but this really isnt the server for that

[late pawn]

And I dont. Im simply sharing my opinion. This is a discussion channel after all

[late blade]

It's quite pedantic if you just point your finger, judging towards a community you are not part-taking, you should join the fun and at least learn to know the people in the community

[late pawn]

Honestly? Im not on discord enough for that kinda thing. I only even check the daily theme because it pings me. Is the sharing of art not for the purpose of being judged and seen by others? As for taking part, im not gonna spend 6 hours drawing something for a theme that changes daily in a community where most would question what the point even is when Ai can do it faster and realistically, better

If the community would like to get to know me, sure, but im not a part of it enough to get to know them

[late blade]

hehe, then turn-off pings if you don't want them, that is also an option, but I would rather have you come down of that arrogance-cloud and be part of the fun down here on earth

[late pawn]

Im not in any kind of arrogance cloud? Im sharing my opinion on the pieces turned in for the daily theme. In a channel whos sole purpose is to share opinions

I even plan on one day taking part if I like a theme enough which is why I keep the ping on

[late blade]

cool, I'm eager to see what you got

I promise not to just react negative because you reacted negative on mine

[late pawn]

By all means criticise it as I criticise others. I like constructive feedback

[late blade]

merit where merit is due

[late pawn]

I wouldnt think that low of you

[brazen elm]

Is there a browser that chatgpt image AI has

[verbal sequoia]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[late blade]

Do you mean <sora.chatgpt.com>?

[brazen elm]

Ohhh

I though that's for videos

[late blade]

you can select images, on the prompt input box, lower left, select image instead of video

[novel ibex]

yo

[upbeat roost]

bro ai images are getting sooo good now

[heavy sapphire]

yes

[worthy cape]

Hello

[quartz owl]

how you generate images with chatgpt

[graceful needle]

is sora 2 allready in the netherlands?

[inner phoenix]

HVCAG1

[late hornet]

prompt: please generate an image of a blue dragon.

If you’re using ChatGPT with image generation enabled (like GPT-4 with DALL·E), just prompt: ‘Create an image of [your idea]’

[late hornet]

@quartz owl: You can find more options for creating images in the channel #image-bot using the /draw [command]

[astral ether]

Kode plise

[lean lark]

Is there a way to force gpt-image to use the full version instead of mini?

[late hornet]

You can’t force it in ChatGPT the “mini” version runs by default for speed and cost. If you’re using the API, set model="gpt-image-1" to use the full version.

.oO(2048px)

[devout chasm]

is someone here good with creating realistic AI images if they have product photo and model photo and merging them into lifestyle photos?

[vapid elk]

afaik, yes
but if i missed something, let me know

[lean lark]

Did openai update their image generating? It's being super restrictive now.

[velvet rampart]

In what way? They changed their policy about a month ago to be more strict.

[sinful furnace]

Hey eveyone

[dim cradle]

I am trying to create a profile pic for a sports better at casino. How do I find good prompts for that

[slate shell]

A desperate man wandering alone through an endless desert, under the scorching midday sun, wearing torn and dusty clothes, sunburned face, cracked lips, cinematic lighting, highly detailed, realistic atmosphere --ar 9:16 --v 6

[left bridge]

Anyone else having issues with ChatGPT not giving you what you ask for for things like product photographs. Up until yesterday I was able to put in my standard prompt with a picture I took and it would delete the background, put it on a white background, do professional lighting and keep my product beautiful. Now... crap. What happened???

[dreamy breach]

that's not how it works.

[crimson heron]

@atomic ore Did I see that you asked for the prompt for this image? There is not a prompt for it -- at least not in the traditional sense that one would think for image generation prompts. I've been thinking about writing up my process for tips/tricks article which would explain how this and other of my images typically are done. #daily-theme message

[atomic ore]

Yes, and that makes sense. Thanks for answering!

[crimson heron]

When I post my process, I'll let you know. I might have time to do it this week.

[hollow ledge]

What the... Why chat gpt generating cartoonish yellow drawing texture images in most of the times

What a failure. Can't your team fix it???

[velvet rampart]

And we have known this since April. If you get these yellowish images, just add ”No yellow tone”.

[limber shoal]

can someone give me a prompt for cinematic lightning along with god complex

[velvet rampart]

Just ask the AI to write a prompt for you. Describe what you want.

[merry pelican]

any updates on image uploads? why are they still gone?!

[lean lark]

I'm able to upload images just fine... are you on mobile or desktop?

[merry pelican]

Oh no! any image I uploaded on or before October 23rd has gone missing since that day. I’ve seen others mention having the same issue. The problem is that when you try to use an image from before that date, it doesn’t load , you have to re-upload it if you still have it saved.

[stoic mulch]

is dal·le out of service? my chatgpt keeps telling me it can’t generate images due to general maintenance jobs

[velvet rampart]

Sora's trash now works.

[velvet rampart]

Well, works with a given value. I can transfer an image to trash and it stays there. But I can't empty the trash.

[late blade]

I had to do that per image in full screen

[left ferry]

Hi

[sullen turtle]

Has there been any update regarding our broken uploads on the original Sora sites so far?

[wild portal]

If i buy Dall-e subscription does it stick to more unified characters? I think we’ve been keeping the art style i wan’t cohesive, but now it’s forgotten the character. Won’t even take its own reference

[rich charm]

What's the difference between gpt-image-1-mini vs using gpt-image-1 with quality set to medium/low?

[abstract trench]

Still hoping for colour in AI image generation, colours in AI are dull ....

[wide zinc]

Oh thats why I can't generate any images atm?

[sly carbon]

how can I post a picture I generated with sora 1

and where?

[odd flame]

how to fix 99% stuck image generation?

[velvet rampart]

Trash option does not work on Sora at the moment. Plus side is that the trash has not returned trashed items back to the main.

[velvet rampart]

[In Professor Farnsworth’s voice]: Good news, everyone. The trash works again on Sora.

[velvet rampart]

Bad news, everyone. The favorites on Sora do not work and they cannot be displayed.

[toxic hamlet]

I don't upload images regularly, but is ChatGPT "gaslighting" me, because I uploaded an image for it to analyze and after a few prompts it said it no longer has access to the uploaded image.
I asked it and it said that it has access only at the time of the upload and any other promp after that doesn't see the image anymore.
Anyone aware of that ?

[velvet rampart]

Just regular hallucinations.

I recommend to go back a few prompts and trying to redo with a different wording.

[toxic hamlet]

Already tried it, but it insists that images are not stored longer than a few prompts for personal safety and memory-optimization reasons.

[velvet rampart]

Start a new chat.

[gaunt tartan]

Hi

[long crater]

By default your images & videos are published on old Sora Explore page. You can go to Settings > General to change it to make them private.

[long crater]

Go to #image-bot to generate image. You can start with /draw

[velvet rampart]

More characters are blocked. Now Nintendo's Mario, Yoshi, Bowser and Pikachu of Pokemon.

[jovial thorn]

so who won the previous daily theme?

[tender arrow]

Scam

[deft musk]

Everyone who gets 10 stars is placed in #hall-of-fame , you can look and see for any day you like!

[verbal arch]

when gpt image 2 frfr

[jovial thorn]

uploaded a photo and it got deleted after 10 seconds.

[wary current]

R

[hollow ledge]

Why not allowing 9:16&16:9 ratio

Only generating 3:2/2:3 even i mentioned in prompt

Chatgpt not capable of generating these ratio!!! It's 2025 end still no change

[slate zinc]

The Dall-e image generation within the GPT creator does not work anymore for weeks. Anyone else?

[gusty bolt]

W chat gpt

[versed birch]

trading my 82m 67

[dim cradle]

Is chatgpt down?

[feral belfry]

anyone had any problems with the new guardrails for image creation? seems pretty excessive now. cant even use it for simple depictions anymore.

[velvet rampart]

The guardrails have been excessive since April. Please, note that they take the whole chat as a context window. If you can’t generate now, try in a new chat. Often this solves the problem.

[compact wraith]

I think the guardrails are coming down in December

Through age verification*. There's news on it

[upbeat crane]

how to send image to daily-theme?

[mild swan]

Gng Chat gpt im sorry bro.. I didnt mean to call you a dirty clanker

[cold forum]

Just making about Sonic, SpongeBob, Mickey Mouse of stories!

[verbal sequoia]

Get notified for all OpenAI announcements!

Visit id:customize to pick up the <@&1408186587606679582> role. Everyone will still be notified for large releases, updates, and events regardless.

[cold forum]

Ok so cool

[winter coral]

Chatgpt isn't best for their images when they flag everything. Plus isn't that good. It has really really strong guardrails

[velvet rampart]

Have you tried to do the requested images in new chats? Since last April, the guardrails have been high. Please note that the whole chat is a context window. Usually images that you try to generate, succeed when do them in a fresh chat.

[winter coral]

Yea. They just say they can't create a person or whatever.

[velvet rampart]

It depends on how and what you request. I just tried to generate a generic Japanese man’s photo, and it came out nicely. Named persons are always iffy, and modifying a person’s photo is also. It also depends on the entire conversation thus far, as well as what you have in the chat. If there is even something strange or risque, the generation might fail. The best way is to test in a new chat. Ask the AI to generate a prompt if you don’t have one ready. Sometimes it is annoying to hunt down what is actually blocking. For example, last February word “rose” caused false positive flag on Dalle3 generation and it blocked the image formation.

[sullen turtle]

Has there been a fix or solution for our broken uploads, besides re-uploading them?

[velvet rampart]

Nope.

[merry pelican]

they need to fix this!

[sullen turtle]

Exactly! I am sure they are aware of this, though.

[elder crescent]

What does "There was an unexpected error running this prompt" mean ?

It counts as cuota usage

[austere sky]

Hey Miwwa! Always copy command & copy paste. Even this can be re-analyzed!

[merry pelican]

as well as all uploaded images!

[limber shoal]

ye man

[ivory epoch]

Alright guys discuss some chatgpt shall we

What good do you think you he new model brings to us

The

[livid oasis]

If someone wants an insane prompt lmk

[velvet rampart]

Good news everyone. Favorites on Sora are back.

[tender viper]

hello

Hello

[carmine magnet]

heh. already had a daguerreotype style. hence spelling it properly just there. 😄

Style:

Hyper-realistic daguerreotype-style scene, 19th-century Victorian formal attire—high-collared coat, cravat, and pocket watch chain—rendered in silver monochrome tones. Hyper-detailed textures with subtle tarnish along the vignetted metal frame; mirror-like reflective surface with faint patina. Stoic expression, shallow depth of field, and dramatic natural side lighting casting soft shadows. Antique photographic atmosphere with faint scratches, fogged edges, and a ghostly aura. Setting blurred into abstraction, evoking a melancholic, timeless elegance.

Subject:

Aspect Ratio:

[cold forum]

Yeah no away

[limber shoal]

funny how chatgpt flags the most appropriate images ever and famous personalities and it still counts as a request

[topaz ocean]

how do you get chatgpt to generate you something withnout him stopping? it keeps stopping and saying that he couldng generate it. I want him to generate a picture from my fav show to simpsons style and he has problems

[deft musk]

Stuff like that is intellectual property, belongs to someone else. Most of that there's rules about, so the model can't make it. ToS tells us, https://openai.com/policies/row-terms-of-use/

We're only suppost to ask for stuff we actually own, so it's often best to create something original with the model.

Your content. You may provide input to the Services (“Input”), and receive output from the Services based on the Input (“Output”). Input and Output are collectively “Content.” You are responsible for Content, including ensuring that it does not violate any applicable law or these Terms. You represent and warrant that you have all rights, licenses, and permissions needed to provide Input to our Services.

[topaz ocean]

ye but then its something i dont like, i dont want create my own characters or something like that, why is it a rule? i really wish i could generate episodes about my fav show. I know disney wouldnt like it, but they can tell its ai generated, always, i do it because i want so much to make my own episodes. I think this restriction is just unfair and gives away the freedom, only people who make OCs wil be happy with that, but i am not into OCs or something that is ''generate a lion'', just not my cup of tea. I prefer stuff like:
''american dragon jake long in movie, real life ''
'' american dragon jake long in star wars''
'' jake long realistic, or anime version''

I see these videos, images on internet and i want to do this as well.
I hope one day, Disney gives permission to openai to generate disney characters.

[deft musk]

What you said last, about companies who have the IP rights giving permission, that's key, yes.

[topaz ocean]

actually, i am just roleplaying with chatgpt about the show, bc character ai deleted my disney characters, but i am sad that the chatgpt wasnt made for roleplaying, so the sectences are always so long and i ask chatgpt to make character ai sctences, but it has really better memory, really better knownledge so i dont need to tell it, but there is one thing.

When i roleplay for hours ( i ws able to roleplay with chatgpt for like 5 d in a row ) then it starts lagging and i must ask chatgpt to write me what was the roleplay about so i can continue in another chat.
and also happens to me, that chatgpt plays as my character, or describes what i am doing or he decides what i will do and its annoying xd even i put a memory '' never play as my character, never talk or describe as me, never ask on end of sectence what i will do next )) but he seems to not follow that.

But I ENJOY ROLEPLAYING with it. So i will buy probably business - or what is worth to buy? if i only want to roleplay mostly and generate pictures?

but if openai recieves permission, i will start making my own episodes

[deft musk]

Buying a membership is up to you. I can help you find the helpfiles that describe the differences if you want to see that; a Business membership requires at least two seats, for example, and has various differences from Plus, which is the one that individuals more often get for themself if it's not for a business with multiple employees using it.

About the model showing what you do, what I would do is copy the output you didn't like, and change it so it's what you do like, show the model what you want so it has a pattern to follow.

[topaz ocean]

oh thats smart, i can try, yea i use the trial plus right now and i enjoy that. I tried to look for other models which would be made for roleplaying

ohhh and question: when does Sora 2 will be aviable in czech republic?

sorry for my english grammar, thats why i use mostly chatgpt to help me

[deft musk]

We only know what OpenAI tells us, when they do, here's the helpfile to watch to see where Sora 2 is currently available: https://help.openai.com/en/articles/12461230-sora-app-and-sora-2-supported-countries

[topaz ocean]

is using VPN allowed?

[deft musk]

VPN is allowed, for the normal purposes of a VPN and not to deceive about where you are using services from.

If you used any method to try and seem to be in a different location that would mean using services not allowed in your actual location, that would break ToS.

[topaz ocean]

I understand.

Also last question, if it comes to generating, chatgpt promised to generate a 3D model for me for blender, would this be sometimes a thing in future?
would help me a tons, because i dont understand blender, i dont know how to model.

[deft musk]

I haven't used agent, https://help.openai.com/en/articles/11752874-chatgpt-agent so I'm not sure if ChatGPT directly can even with agent do that sort of thing.

However, you can explore trial and error of it making stuff for you to use in a program you have installed and running. If there's an error message you can show that to ChatGPT and it can discuss what to change.

You can also ask it if you choose to help you learn how to model.

[topaz ocean]

WAIT, it can literally open blender in the agent mode? i didnt know xdddd i used it for like '''chatgpt, open chatgpt and text with him and force him to believe you are ai'' but didnt know it can have such uses XD

and sure, i dont have any uses now but if i buy another plus, i will definitely try

[deft musk]

I have not used agent. So I don't know if it can specifically use Blender at all.

[topaz ocean]

ah mb, but i can try download it in his virtual browser if it will work

[lime sapphire]

How do I, like, start making my own pictures with AI from my own pc?
i don't want to use Sora or ChatGPT, etc.; I just want to have my own ai that makes pictures like how sora and ChatGPT do.

But idk where to start and most people i met who makes their own without chatgbt for an example never tells you where to start

[vapid elk]

try ComfyUI or AUTOMATIC1111, other than that, ask more on #ai-discussions since other channels are only for OpenAI related stuff

[dim cradle]

Can I post images here along with questions?

[deft musk]

I don't think we can post images in this channel, but #1155772063596953642 or #images-canvas might work!

[prime nova]

hi all, any idea about how to make videos with ai tools in garments, the idea is to make a video presenting a garment with a model walking and presenting it for 10-15 seconds? tx for your help

[vapid elk]

should be simple enough to do with Sora with a very simple prompt

just provide a picture of the garment alongside the prompt describing what you want and it should work fine

[cold forum]

Story of the scenes

[river citrus]

I wanna be deadass here i am actually sick of the damn content policy sometimes

may i vent about it
(Even just for 5 seconds)

[cold forum]

Story of the scenes

[plucky hare]

Hey y'all, this channel is for discussing OpenAI image models - please feel free to discuss non-OpenAI things in the #ai-discussions channel. Removed just a couple of messages. Thank you!

[late blade]

@mint cradle please don't post images done by gemini in the daily theme

[mint cradle]

Oh ok i thought it was free, sorry

[velvet rampart]

Bad news everyone. At the moment, Sora’s My Media shows only latest generated images and failed generations. The generations are now gone from the feed but they can be accessed via favorites and search.

[orchid current]

Trying to make characters, is the image generator even good enough to make a character?

[crimson pendant]

How do you get ChatGPT to edit a image not generate a picture of the example image

[small geyser]

have you found a solution to this ?

[chilly spade]

No

[sullen turtle]

In my experience, it does work well enough to create Consistent characters. What I find works is to upload an image of the character I want to use and then make sure that my prompts contain details that will help the character remain consistent. You can also upload specific clothing and accessories and in the prompt, mention that the character is wearing that clothing or accessory according to the reference photo.

[late blade]

Currently unable to access renders I've done in the past week or so

including the ones I've done today

[velvet rampart]

You can access them via the activity button on the right hand corner, search function or favorites.

[late blade]

Yeah, still frontend UI is not consistent

[velvet rampart]

Fix one thing, break an other since October.

[ocean crater]

“A cruel middle-aged man standing outside his small dusty hut in a rural Indian village. He is thin, slightly hunched, with messy hair, wrinkled angry face, and wearing an old brown kurta and loose pajama. In front of him stands a strange magical chicken glowing faint golden light. Clean blue sky with a few soft clouds. Bright daylight. Full-body characters clearly visible. Hyper-detailed 3D animation style like Pixar/DreamWorks. Vertical 9:16, 1024×1792

[frozen copper]

Any plans for OpenAI to do massive upgrades to image gen to compete with Nano Banana Pro? Cause their latest release ourperforms GPT5.1 in that area

[late blade]

The plan is always there, the release schedule is as soon as it's ready

[velvet rampart]

Am I halluciating or is the image quality slightly higher now? Less user, less strain on systems?

[late blade]

same for me

[velvet rampart]

And on Monday we are back where we were. It is nice to get higher quality sometimes.

[late blade]

@mint cradle can you stop posting images that are not done with OpenAI models? You got another SynthID positive post by a google model

[foggy kestrel]

is it just me or the new update make the quality of the image look very bad?

why are you touching the image if no one complained? why do you have to ruin it

[oak abyss]

Image generation under browser is taking too long. Anyone else experiencing slow performance with image generation tools in browser-based applications?

[velvet rampart]

Could you please provide an example and tell more how? I just test three ways the image generation and do not see quality drop.

[late blade]

Getting incomplete renders again