Allow initial_noise_multiplier to be scaled to zero

[prescience-data]

Is there an existing issue for this?

• I have searched the existing issues and checked the recent builds/commits

What happened?

https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/a3ddf464a2ed24c999f67ddfef7969f8291567be/modules/shared.py#L420C105-L420C105

This line enforces a lower limit of 0.5 where people have reported better results / less color distortion by setting values of ~0.2 - ~0.3.

There doesn't appear to be any reason for this artificial 0.5 limit that I can find in any existing issues.

Steps to reproduce the problem

1. Run multi stage img2img inpainting tasks with 0.5
2. Run same img2img inpainting tasks with 0.25
3. Lower values in some case (ie photography) results in less distortion.

What should have happened?

The slider should be allowed to scale from 0 - 1

Version or Commit where the problem happens

https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/a3ddf464a2ed24c999f67ddfef7969f8291567be/modules/shared.py#L420C105-L420C105

What Python version are you running on ?

Python 3.10.x

What platforms do you use to access the UI ?

Windows

What device are you running WebUI on?

Nvidia GPUs (RTX 20 above)

Cross attention optimization

Automatic

What browsers do you use to access the UI ?

Microsoft Edge

Command Line Arguments

Not relevant

List of extensions

Not relevant

Console logs

Not relevant

Additional information

No response

[catboxanon]

If this is changed I think it should actually be revisited before doing so. The implementation (#5373) was meant to match NovelAI's and per #5351 (comment) it doesn't. I might take the time to look more into this.

[nbalive2k]

yes, it would be very useful
I use in collab
I have to change the values in shared.py manually all the time