stable-diffusion-webui
stable-diffusion-webui copied to clipboard
Missing samplers/Upscaler from txt2img Highres fix[Bug]:
Is there an existing issue for this?
- [X] I have searched the existing issues and checked the recent builds/commits
What happened?
When opening the default WebUI, many samplers/Upscalers are missing from the automatic upscaling function (Highres fix) on txt2img.
The attached pictures which upscales are available in txt2img and img2img. My favorite upscaler is LMS Karras, but it is not possible to select that on txt2img.
Help?
Steps to reproduce the problem
- Open WebUI
- Check Highres fix
What should have happened?
All samplers should be available independent of which tab/dropdown is selected
Commit where the problem happens
COLAB
What platforms do you use to access the UI ?
Other/Cloud
What browsers do you use to access the UI ?
Google Chrome
Command Line Arguments
No
List of extensions
Console logs
N/A
Additional information
Please help
Unless I've been badly misunderstanding something, the sampler and the upscaler are two totally separate things.
The way 'hires fix' works currently is:
- Creates first image with selected Sampler.
- Creates a temporary, scaled-up image with the selected Upscaler.
- Creates the final image by doing an img2img operation on the larger temp image, using the same Sampler as in step 1.
So, if you are selecting LMS Karras as the original sampler, it will use it again on the final image.
It's interesting to see the differences between using, for example, 'Latent' and 'None' as the upscaler. If the final image is changing too wildly from the original, you may need to lower the Denoising strength setting a little (it's situational), particularly for 'Latent', in my limited experience.
Unless I've been badly misunderstanding something, the sampler and the upscaler are two totally separate things.
The way 'hires fix' works currently is:
- Creates first image with selected Sampler.
- Creates a temporary, scaled-up image with the selected Upscaler.
- Creates the final image by doing an img2img operation on the larger temp image, using the same Sampler as in step 1.
So, if you are selecting LMS Karras as the original sampler, it will use it again on the final image.
It's interesting to see the differences between using, for example, 'Latent' and 'None' as the upscaler. If the final image is changing too wildly from the original, you may need to lower the Denoising strength setting a little (it's situational), particularly for 'Latent', in my limited experience.
Oh .. that changes my understanding. Thanks
Done