Type | |
Stats | 465 0 |
Reviews | (49) |
Published | Jul 18, 2024 |
Base Model | |
Usage Tips | Clip Skip: 2 |
Hash | AutoV2 045906A063 |
Tis is a SD1.5 LCM Model
This model was supposed to be The Anime WonderMix V6 but... It turned out to be a bit more NSFW than usual.
But still can easily produce beautiful SFW image.
I played a lot with it, so you will see different output : it's a mix of different Style tags. just check the generation prompt when an output catch your eyes ;)
Details and generation settings
Go on The WonderMix page and read the V14 part. It's the same settings, and, the character list work also on this model (yup no LoRa needed for thoses)
Just one thing, if you want to use a specific style tags (Pastel color, anime coloring, an artist etc...) put them at the end of the prompt.
The baked VAE is z-vae
[Edit]
With a LCM Model, your CFG sholud always be around 1 (between 1 and 1.5), even if you use another Sampler/Scheduler.
Good other Sampler/Scheduler to use with a LCM Model are the Euler with SGM Uniform or Karras but, avoid the DPM.
[Edit 2]
on A1111, don't do an HiresFix or use ADetailer with another sampler than LCM..... It will turn...Well.
On ConfyUI, just don't use a latent or pixel upscale before your second kSampler (i know it defeat a bit the prupose... But you'll still get a denoise)
Merged Model
yeah, only model were used.
Credit goes to thoses creator, and thanks to them
Have fun, and don't hesitate to post your generation !) (i found a lot of things by looking to other generation metadate, thanks a lot!)