Prompt in image file

NovelAI embeds the prompt and other parameters in the PNG file, and other softwares may also do this in future. Should we tag those files as 'has_prompt'? Currently we can use a script to automatically detect this.

Also I think we should encourage everyone to upload the prompt along with the image, and with that we can tag the images only according to what is actually in the image, rather than what is in the prompt.

Not sure if the filename counts, as my understanding has been prompts are encouraged to be added as commentary on the post or as a comment. If it has what was used to generate the image, then I don't see why not.

As for your other comment, I haven't seen people favoring tagging based off of prompts over the content of the post itself. It's less been an "encouragement" and more so the expected standard.

The work around as we don't have a special field for prompt now is the artist commentary.

Do all the tags on the filename don't get cutoff for sure? I guess it depends on the software like novelai vs others.

It's not the filename, but if you open the PNG file using notebook (or your favorite text editor), you'll see texts like
'tEXtTitle AI generated imageŽñR£

tEXtDescription 1girl, crowd, exhibitionism, public_indecency, female_orgasm, revealing_clothes, ribbon, bdsm, nipple_clamps, cum_on_body, cum_on_clothes, cum_on_face, after_sex, skindentation, fishnets, outdoors, day, street, looking_at_viewer, selfie, v_signpÄPp tEXtSoftware NovelAI…]­ß tEXtSource Stable Diffusion 81274D13Á)¹h
…tEXtComment {"steps": 28, "sampler": "k_euler_ancestral", "seed": 3757612401, "strength": 0.69, "noise": 0.667, "scale": 11.0, "uc": "lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad feet, futa, futanari, yaoi, huge_breasts, large_breasts"}'
in that case I think we can directly tag it as 'has_prompt', or use a script to automatically add the prompt into the commentary.

Ah, we can automate that in the future yeah.
Many places strip metadata tho.

https://aibooru.online/posts?tags=exif%3APNG%3AComment+exif%3APNG%3ADescription+
https://aibooru.online/media_assets/1567
for now you can go to those pics that have it and look up by clicking the arrow next to the image resolution

Thanks bro

Another common tEXt chunk with prompts is exif:PNG:Parameters (used by the self-hosted web UI [1][2]), to search for both options you can use (exif:PNG:Comment exif:PNG:Description) or exif:PNG:Parameters.

I just noticed that you put some instructions in the upload page. When I already have those metadata in the file, should I copy-paste them into the comment?

laji said:

I just noticed that you put some instructions in the upload page. When I already have those metadata in the file, should I copy-paste them into the comment?

If you make sure it is really here in the metadata then its ok for now, we will soon migrate them to appear on the post page itself.

Prompts are not being parsed from formats other than PNG, where, for instance, the SD WebUI embeds the prompt in JPGs as, I believe, the exif tag UserComment [1]

edit: one day after this post, on a new upload the page autoamtically fetched the prompt from my JPG, so I was wrong, or this was just changed?