stable-diffusion-webui
stable-diffusion-webui copied to clipboard
Feature: validate settings in UI before starting generation
Is there an existing issue for this?
- [X] I have searched the existing issues and checked the recent builds/commits
What would your feature do ?
Some combination of he just don’t make sense. For example running a batch generate of 100 batches with a fixed, non random seed is just a waste of energy
this happens to me a few times already:
I boot my pc and want to continue where I left off with my last prompt. What I usually do is open the output folder, grab the newest picture and load it to the png info tab and click the “send to txt2img” button
The I go to the generation settings tab, make some adjustments to the prompt and increase the batch size to 100 because I always want many variants to choose from
Often times I forget to change the seed to -1 again which was set to the last images seed by the send to txt2img and instead just impatiently press the generate button
so the exact same image would be generated 100 times in a row which does not make any sense
Proposed workflow
When batch size is set to a value that is not -1 while the batch size and:or batch count is larger than 1, display and error message overlay over the generate button and don’t start generating until either batch size/count are both zero or seed is -1
Additional information
No response
For example running a batch generate of 100 batches with a fixed, non random seed is just a waste of energy
true, but the issue is that we cannot know that the some parameter won't be change by some other thing like an extension or script for instance XYZ grid script, you can set the seed in script, and if you do the seed in your initial input does not matter and so we cannot make the prediction base on the ui input that this will be a waste of energy
also if you set batch count to 100, the seed should increment from batch to batch so this should not be an issue
Often times I forget to change the seed to -1 again which was set to the last images seed by the send to txt2img and instead just impatiently press the generate button
if you don't want to reuse the seed at all then you could just set the UI to discard the seed when copy and pasting
So the condition would have to be seed not equal to -1 and batch size / batch count both equal 1 and no script provided
So the condition would have to be seed not equal to -1 and batch size / batch count both equal 1 and no script provided
the condition you provide will block generating images with set seed at batch size and count 1
also you're not factoring extensions as extension also can change seeds if you exclude built-in ones most people will have at least a couple of third-party extensions so this basically makes this condition useless
if you can make the assumption that if all the inputs are exactly the same the output will be the same and you want to block this type of generation then you not only have to compare all the inputs but also all settings and possibly some other internal values
and then the issue if there are two users who somehow decides to enter the same input if this is implemented the first user would get an up but the second user won't so now you're going to have to test for the user somehow
if you're doing img2img then you have to check the input image as the same, if you're inpaint you also have to check the mask not to mentioning some extensions can change the mask
the complexity just keep scaling
if there's anything that can change the output it has to be factor in
User error.