feat: DALL-E Compatibility Improvements #581

Open
opened 2025-11-11 14:26:42 -06:00 by GiteaMirror · 1 comment
Owner

Originally created by @tusing on GitHub (Apr 2, 2024).

Some feature requests from issues I've run into during my own usage of DALL-E via this project:

  • API feature implementation
    • DALL-E 3's API has two image quality options: standard and hd, but there is no way to select which one is used.
    • We should support multiple image generation for DALL-E 2/3.
      • For DALL-E 2, the n parameter (1≤n≤10) controls how many images can be generated.
      • For DALL-E 3, the n parameter must be 1, but we can request more images in parallel.
  • User-experience related
    • The DALL-E 2/3 API does not take a steps parameter, but it is visible in the UI, pre-set to 50, which may cause confusion.
    • DALL-E 3 can only take 3 resolutions (1024x1024, 1024x1792, or 1792x1024) but the resolution input is free-form, which may cause confusion.
    • DALL-E settings are not preserved, and there does not seem to be a way to preserve them between Docker container restarts. The API key and API settings are both wiped, even when passing in the API key environment variable and ensuring the volume is mounted correctly.
  • Prompt-revision related
    • DALL-E 3 "automatically re-writes the prompts for safety reasons" but it is possible to disable this via the string provided in the API guidelines. We could make this an option.
    • We should show what the prompt used for the image was, which is provided by the API response.

Related: DALL-E API Documentation

Originally created by @tusing on GitHub (Apr 2, 2024). Some feature requests from issues I've run into during my own usage of DALL-E via this project: - API feature implementation - [ ] DALL-E 3's API has two image quality options: `standard` and `hd`, but there is no way to select which one is used. - We should support multiple image generation for DALL-E 2/3. - [ ] For DALL-E 2, the `n` parameter (`1≤n≤10`) controls how many images can be generated. - [ ] For DALL-E 3, the `n` parameter must be `1`, but we can request more images in parallel. - User-experience related - [ ] The DALL-E 2/3 API does not take a `steps` parameter, but it is visible in the UI, pre-set to 50, which may cause confusion. - [ ] DALL-E 3 can only take 3 resolutions (`1024x1024`, `1024x1792`, or `1792x1024`) but the resolution input is free-form, which may cause confusion. - [ ] DALL-E settings are not preserved, and there does not seem to be a way to preserve them between Docker container restarts. The API key and API settings are both wiped, even when passing in the API key environment variable and ensuring the volume is mounted correctly. - Prompt-revision related - [ ] DALL-E 3 "automatically re-writes the prompts for safety reasons" but it is possible to disable this via the string provided in the API guidelines. We could make this an option. - [ ] We should show what the prompt used for the image was, which is provided by the API response. Related: [DALL-E API Documentation](https://platform.openai.com/docs/guides/images?context=node)
GiteaMirror added the help wanted label 2025-11-11 14:26:42 -06:00
Author
Owner

@JOduMonT commented on GitHub (Jun 6, 2024):

I was about to say something ;)
For now Open-WebUI only support 512x512 and 1024x1024 repectively for DALL-E 2 and DALL-E 3.

  • but does'nt support DALL-E 3 in size like 1024×1792 or 1792x1024, nor HD
@JOduMonT commented on GitHub (Jun 6, 2024): I was about to say something ;) For now Open-WebUI only support 512x512 and 1024x1024 repectively for DALL-E 2 and DALL-E 3. - but does'nt support DALL-E 3 in size like **1024×1792** or **1792x1024**, nor **HD**
Sign in to join this conversation.
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/open-webui#581