AutoGPT icon indicating copy to clipboard operation
AutoGPT copied to clipboard

Support image_size for generate_image command

Open JeremyFabrikapp opened this issue 1 year ago • 1 comments

Duplicates

  • [X] I have searched the existing issues

Summary 💡

Command Generate Image "generate_image" is a great addition. However the image size is hardcoded, I propose the issue to make it controllable with an optional argument image_size.

Pitfalls :

  • Dalle has only 3 sizes "256x256", "512x512", "1024x1024".
  • StableDiffusion works with multiple of 8 as valid size, but image may suffer loss of context > 512px.
  • image_size argument should be optional and not make it more complicated to generate an image without specifying the size.

Examples 🌈

  • Stable Diffusion Replica, size can be changed https://replicate.com/stability-ai/stable-diffusion

  • Dall-e API https://platform.openai.com/docs/api-reference/images/create#images/create-size

Motivation 🔦

This will help the user to generate an image at the needed size.

I will push a PR in that direction.

JeremyFabrikapp avatar Apr 08 '23 14:04 JeremyFabrikapp

This can be closed by https://github.com/Significant-Gravitas/Auto-GPT/pull/1477 right?

edcohen08 avatar Apr 25 '23 03:04 edcohen08

Yes @edcohen08 it's been implemented I'll close the issue !

JeremyFabrikapp avatar May 02 '23 00:05 JeremyFabrikapp