Prompt/settings - read them from generated metadata, save and use them as presets

[HanJammer]

I knot this feature is partially integrated as the styles.

I would love the option to open previously generated txt file with the UI and boom - you have all the previously saved settings which allows you to further enhance/regenerate your prompt. Also it would be cool to have the ability to save the UI settings in the independent presets (it would save the prompts and setting directly in the UI for the quick access).

Also IMO it's way better (and more portable) to keep the metadata in yaml (like hlky's UI does) instead of txt file (like AUTOMATIC1111 UI does currently).

[ClashSAN]

Could you describe what that UI button does? I never figured it out and its been out a while.

Also IMO it's way better (and more portable) to keep the metadata in yaml (like hlky's UI does) instead of txt file (like AUTOMATIC1111 UI does currently)

I thought all metadata is stored in the png by default. Then you can enable saving metadata to text files additionally.

[HanJammer]

Create / apply style as far as I know simply saves and restores the prompt... nothing more...
Yes, metadata lands in png... unless you are using jpg-s. But metadata stored separately is the way to do it (that's how professional digital archive/library systems do it). Embedding it in the images is not that good idea - when sharing images on the facebook or reddit it will be removed, and having metadata in the separate file makes it super easy to post and share (just copy/paste).

Also it's true that there is a PNG Info button allowing you to read metadata from PNG but like I said it would be most useful to do the same for the txt (or better - yaml) metadata files.

[mattehicks]

I wanted to follow up here, were you able to find some resources for this? I'm trying to run img2img in batch mode, using metadata for the prompts. I haven't found any way to do this yet.

[noantidote]

just a follow up here. were you able to get metadata for img2img? i'm trying to write a script to scrape all prompts from png files in a folder and print out a csv file of prompts. I need more info on how the "prompt" data is keyed in the png file. do you know?

[missionfloyd]

@noantidote Parameters are read like this.

from PIL import Image

image = Image.open("image.png")
parameters = image.info["parameters"]

Returns a string. It can be parsed to a dict with parse_generation_parameters()

[noantidote]

thankyou very much.

…

On Sat, Sep 23, 2023 at 8:32 PM missionfloyd ***@***.***> wrote: @noantidote <https://github.com/noantidote> Parameters are read like this. from PIL import Image image = Image.open("image.png")parameters = image.info["parameters"] Returns a string. It can be parsed to a dict with parse_generation_parameters() <https://github.com/AUTOMATIC1111/stable-diffusion-webui/blob/5ef669de080814067961f28357256e8fe27544f4/modules/generation_parameters_copypaste.py#L210-L316> — Reply to this email directly, view it on GitHub <#1737 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3RW2ZR63R5OAGNVSNLLYK3X36ES3ANCNFSM6AAAAAAQ5X6SLQ> . You are receiving this because you were mentioned.Message ID: <AUTOMATIC1111/stable-diffusion-webui/repo-discussions/1737/comments/7091479 @github.com>

[ewebgh33]

I posted today about this before I knew about this thread.
A good way of doing this would be to extending functionality and this can be done without changing the current default behaviour.

Currently style pulls in a prompt and/or negative prompt.

Cool, keep this. But from now on saving prompt styles would also save other data (like seed, CFG) and you could (with a checkbox next to the styles dropdown) pull in that additional data too.
I posted here:
#10725

This would help us test prompts (with seeds) in an XY plot against different checkpoints or loras.

[noantidote]

very good news. thankyou

…

On Sun, Sep 24, 2023 at 10:03 PM EmmaWebGH ***@***.***> wrote: I posted today about this before I knew about this thread. It would adding functionality to saved prompt styles - extending functionality. Default behaviour when reading in saved styles would not change, but from now on saving prompt styles would also save the other data (like seed, CFG) and you could (with a checkbox next to the styles dropdown) pull in that additional data too. #10725 <#10725> This would help us test prompts (with seeds) in an XY plot against different checkpoints or loras. — Reply to this email directly, view it on GitHub <#1737 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A3RW2ZQBM6WWXKWVYI6TKG3X4DYA3ANCNFSM6AAAAAAQ5X6SLQ> . You are receiving this because you were mentioned.Message ID: <AUTOMATIC1111/stable-diffusion-webui/repo-discussions/1737/comments/7096920 @github.com>