Additional Info
- Fiddled with the block weights.
- Lora support but still a bit finicky.
- NSFW capability still decent, even with Animagine XL added.
- Alpha block weights tweaked differently from 'Final'
- Beta block weights shared with Final
- VAE is baked in from the start, which is v1
Description
Model Details.
V1-Tests
- r_ain for SnowPony
- Goofy_Ai for Shiny Nai style for Pony
- Goofy_Ai for Floox Style PDXL
- Minthybasis for 4th tail
V2-Test
Redid the merge and removed the baked-in loras for a cleaner sheet. Added some bits and piece as well then played with the weights, gone with sum twice as I was merging two models at a time.
V2-Test Shiny?
(TBA)
Planning to bake another lora into this but will release as a separate file. Less the shine and stuff.
Recommended settings.
it's recommended to use a lower classifier-free guidance (CFG Scale) of around 6 - 7, sampling steps between 20 and 26, and to use Euler Ancestral(Euler A) as a sampler.
Suggested prompts: "score_9, score_8_up, score_7_up,"
Suggested negative prompts: "score_4, score_3, score_2, score_1, ugly, bad hand, bad feet, extra feet, lowres, (bad), text, error, fewer, extra, missing, worst quality, jpeg artifacts, low quality, watermark, unfinished, displeasing, oldest, early, chromatic aberration, signature, extra digits, artistic error, username, scan, abstract"
Notes
Based on SnowPony and 4th tail with Shiny Nai style and Floox Style PDXL baked in with different weights per testing. I found that 0.5 weight looks the best according to my eyes (sorry, my eyes are bad). All credits goes to the original authors of the models and resources used for this merge.
Follow me on Pixiv