Super Working.
Images are of good quality.
Some images have error so I Retry that prompt and get better image.

New Update ,New Features:

1. Promptifier is much better than before which lower the mistake in image.
2. Model is changed from Mixtrail-8x-7B-instruct-0.1 to Nous Hermes Mixtrail DPO
   (2 main reason for changing this is Instruct 0.1 is mainly for text generation so he have low creativity whereas Nous Mixtrail is good has good greativity like chatGPT because it is fine tune on ChatGPT's answers)
3. Model do not ADD DETAILS in FULLY DETAILED prompts. (Suggested by @SvCy )
4. From now more ideas images are not generated till user says AI. (Suggested by @SvCy )
5. From now you can help in making model better.
   .........................................................................................................................
   So, how you can help in making Model Better
   Just Share Your Best and worst generation (Mainly worst so I can reduce mistakes made by bot)
   

And tell me what features you want in future.
..........................................................................................................................
About Future updates of this bot

1. Every image must be private (This features help to protect your generation from showing on pollination's home page)
2. More advanced Promptifier (I find that good quality prompts lead to good quality image, So i want to make users prompts much better for ai to understand).
3. More related Result to your expectation (Prompt) (That's the reason i am asking for your worst generation)
4. Every image must be unique. (Currently when you give same prompt to bot there is 1/10 chances that he generates same image again so to reduce this to 1 in Trillion by using random seeds for each image)
5. Internet Access (Help modal to generate detailed images of unknown person, thing, place, etc.)
6. Faster generation by removing unuseful components from system prompt (Currently have 6000 try to reduce it to 1000).
7. Don't forget to tell me what features you want in future.
   ..........................................................................................................................

6. Faster generation by removing unuseful components from system prompt (Currently have 6000 try to reduce it to 1000).

Don't add so many cases, just tell it how to construct the URL with addition Quality prompt,It should be enough to use string splicing. LLMs can do a well in this .

Randomness can be solved through Temp and Top_K. At the same time, I still only recommend Command-R+ - its knowledge base is up-to-date, that is to say, its internal knowledge contains "what the stable diffusion webui prompt looks like", which can reduce many cases.

New Update, New Features:

1. Every image must be private (This features help to protect your generation from showing on pollination's home page)
2. More advanced Promptifier (I find that good quality prompts lead to good quality image).
3. More related Result to your expectation.
4. Every image must be unique. (By using random seeds for each image).
5. Removed unwanted cases. (Suggested by @Mindires ) (Help in making Unique and Faster Generation).
6. Added prompt Grader by @Thepickledegg . (It helps the promptifier in classifying prompts. So, Model generate Different number of images for different quality of prompt.)
7. Model shifted to LLAMA 3. (Because it is better than Nous Hermes Mixtral DPO).
   ..........................................................................................................................

Future Features:

1. Internet Access (Help modal to generate detailed images of unknown person, thing, place, etc.)
   (I attempted to include this in the current update but encountered numerous problems. Therefore, I have decided to address these in the next update.)
2. Shifting Model to Command R+ (Suggested by @Mindires ) (I am attempting this, but currently, Command R+ has many issues. Therefore, the model will be shifted once all issues are resolved.)
3. Suggest features you want in future.
   ..........................................................................................................................

Bot link -> https://hf.co/chat/assistant/6612cb237c1e770b75c5ebad

Great job

Create for me a portrait of a 1952 Pegaso Z102 in a clean and colorful palm forest, uneven light shadows and raw style.

@Tamer12 you should write that on the assistant link shared here maybe & tweak further based on results

@Tamer12 Here are some results for u -> https://hf.co/chat/r/OxAt45i

Please try yourself -> https://hf.co/chat/assistant/6612cb237c1e770b75c5ebad

This is so great!
I got a problem, after several messages I get this error:

Input validation error: inputs tokens + max_new_tokens must be <= 8192. Given: 6543 inputs tokens and 2047 max_new_tokens

Starting another chat with this bot just gave a generic error and couldn't complete. Anther unrelated chats works fine.

Hey @iapdem , just refresh the page and start a new chat with the bot.

Thanks, it helped.
Actually I was trying to build on the first prompt, adding more and more detiles as we go along (same as I do when I coding, for example), is this workflow not supported?

Actually I was trying to build on the first prompt, adding more and more detiles as we go along (same as I do when I coding, for example), is this workflow not supported?

Actually, HuggingChat has limitation for every Model.

Got it.
And thanks again for this fine model!!

Uhm, what exactly does the LLM do? The picture is produced by pollinations.ai with a REST API:
https://image.pollinations.ai/prompt/Photorealistic%20image%20of%20Monkey%20exploring%20Moon's%20craters%20and%20mountains%2C%20with%20a%20spacesuit%20on%2C%20and%20a%20futuristic%2C%20sci-fi%20inspired%20background.?width=1024&height=1024&nologo=poll&nofeed=yes&seed=85432109

@enzomich Try giving him simple prompt he will generate best image according to it by improving prompt quality for AI, So, these helps non-techy to generate best images.

I said, "kazuya Mishima shooting a pistol" and it created a female in one of the images.

@Zainmdali
Issue Resolved
Check This - https://hf.co/chat/r/M0RVeEo

An abstract scene filled with interwoven vibrant and dark colors representing the complexities of life. Multiple figures with exaggerated expressions and gestures depict arguments, disagreements, and various biases. The figures face different directions, symbolizing divergence and lack of unity. Symbols like a broken mirror, pierced heart, and tangled lines are scattered across the scene, representing constant conflicts and prejudices. Use stark color contrasts, such as red and black, to highlight emotional intensity, with light and shadow emphasizing key elements. The background features chaotic patterns or mist, adding to the sense of life’s intricacies

https://hf.co/chat/r/xAtSvfd
My prompt was: a black pigeon flying past a dragon, and it only created the pigeon.

İ write Albert wesker vs tyrant but output Have that pics

Hi, we prefer if people post new assistants on the community discussions here to avoid cluttering the feed: https://huggingface.co/spaces/huggingchat/chat-ui/discussions/356

Not working?

why is the old image generator model not working? is it gone forever?

Thank you

หมา