Hello everyone,
It’s been a while since my last upload! The reason for that is that I’ve been working on a 24-episode anime where the characters change their appearance and outfits from one episode to the next 🤣🤣🤣 — which didn’t exactly make things easy for me!
Our next anime will be Is the Order a Rabbit? (Gochuumon wa Usagi Desu ka?), and next character is the other main actor: Chino Kafū
Sorry for spamming new versions - this should be the final version.
Ver2: refined tags, changing character tag to match the danbooru tag (proposed by another user).
Ver3. even more refined caption (generalization for background-tags, removal of specific exterior an interior locations (such as "forest", "montain", "ocean" and "cafe", "school", "bedroom"). Removal of generic tag like "holding" and "solo"
Note: Filenames of Version 1 (has "ver2" in name) and version 2 (no version info in file name) are misleading - do not mind (you probably should use Ver 3 from now on anyway)
Some outfits might be a bit unstable because the training data is either limited or low in quality — and in some cases, other outfits might have too much influence.
Prompt: look at "Trigger Words" on the right side 👉
Changelog: look at "About this Version" on the right side 👉
quality tags I use:
oekaki, anime coloring, anime screencap, masterpiece, best quality, amazing quality
My Settings (forge)*:
*note, the example images only use high-res fix.
Checkpoint: waiNSFWIllustrious_v140 waiNSFWIllustrious_v150
Sampler: Euler a
Scheduler: Automatic
Samplings Steps: 30 - at least use more than 15! I often see people generate with 8 steps. This it not enough xD (I mean, you should generate you pics the way you want, but the results should be better with more steps)
Hires.fix: R-ERSGAN 4x+Anime6B, 20 Hires steps, Denoising Strength: 0.3-0.5
Extensions (nice to have for little more Details):
Adetailer (highly recommended)
CFGRescale Integrated (forge)
SDXL Styles (anime)
CD Tuner
Detail Daemon
DynamicThresholding
FreeU
SelfAttentionGuidance
Description
final version (hopefully). Refined Captioning even more (generalized background, removed generic tag like "holding", removed "solo" tag)
FAQ
Comments (6)
can u change the tag u write from "chino kafū" to "kafuu_chino" to match the danbooru one? it might help to reinforce in outfits if the model has danbooru data, as wai does
same for the rest of the models (though in the case of mocha at least no extra data will come from danbooru, but to keep consistency)
Hello, simply changing the keywords is not possible - i have to re-train then. I could do a version 2, but i dont think that adding the danbooru tag will benefit the model as much, cause the "main" outfit - the rabbit hose uniform already using the danbooru tag. Maybe I am wrong so ill test this: I will do a model with the tag replaced and then i try to make a model with both tags, thanks for your proposal
Added Version 2, it felt a bit cleaner (although I also did a bit of cleanup in Captioning)
If you download v2, the file appears to be v1 and if you download v1, the file name appears to be v2.
Either the file name is wrong or the file is reversed?
Ahh, sorry my first uploaded file was actually ver2 (the original ver1 was unpublished because of some errors). As for the ver2 i simply overwrote this faulty version. Short: Its okay, but the filename is misleading... fyi there will be a ver3 (for this model only) this evening, cause i changed some other things in captioning like generalization of background, removal of the "solo"... sorry for the Confusion
@IcoNx937 I see, I understand now. Thank you.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.