InvokeAI icon indicating copy to clipboard operation
InvokeAI copied to clipboard

[enhancement]: Magic Prompts

Open xaroth8088 opened this issue 1 year ago • 1 comments

Is there an existing issue for this?

  • [X] I have searched the existing issues

Contact Details

No response

What should this feature add?

The ability to use an LLM (e.g. LLama) to take in a short prompt and expound on it to enhance generated image quality.

Alternatives

Running something like GPT4All in parallel, or using ChatGPT/Claude/etc. to generate the enhanced prompts.

Additional Content

ChatGPT/DALL-E 3 does this secretly in the background (though you can get at the actual prompt after the image is generated), and Ideogram has a toggle for this.

Natural follow-on features would include:

  • Editing the LLM prompt template that's used to generate the enhanced prompt
  • LLM model management for the prompt generator, to pull down any compatible model

xaroth8088 avatar Aug 05 '24 04:08 xaroth8088

Yeah, I use similar in some of my Comfy workflows and it's super cool for creative purposes. For Forge and now Invoke I'm typing manually into LM Studio which is of course somewhat slow...

Invoke is great btw - thanks for all the hard work so far!

stripealipe avatar Oct 02 '24 08:10 stripealipe

There are also some specialized models for this like https://huggingface.co/gokaygokay/Flux-Prompt-Enhance

dennis-rall avatar Nov 19 '24 07:11 dennis-rall