This is a Checkpoint i started with merging with my 3 favorite Checkpoints. Dark Sushi Mix, ReV Animated, Pika's Animated Mix. Other checkpoints can be found in each version detail.
I have been getting some incredible results with this merge and thought that it would be worth sharing with others. It is great at generating both NSFW and SFW content. Would love to see your results :D
Check my other models if you liked this one :)
Hugging Face: https://huggingface.co/BixBit11/DarkRevPikas/tree/main
My recommendations is as follows:
Sampling Method: Euler A or DPM++ SDE Karras
(My personal favorite is Euler A)
Sampling steps: 30-40
CFG Scale: 6-8
CLIP: 2
Hires.fix: upscale by: 1.5x -- Hires steps: 10 -- Denoising strength: 0.3
VAE: I strongly recommend to use this model with vae-ft-mse-840000-ema-pruned
For a more realistic image use:
photorealistic, (hyperrealistic:1.2), beautiful, masterpiece, best quality, extremely detailed face, perfect lighting,
For less realism use:
beautiful, masterpiece, best quality, extremely detailed face, perfect lighting,
My favorite negative prompt:
(worst quality, low quality:1.4), (monochrome), zombie, watermark, username,patreon username, patreon logo, (extra fingers, deformed hands, polydactyl:1.5),
Description
Version 3
This pruned version is a little better at creating 2.5D images. Between the previous V2 i liked V3 better. You can still get more 2D like images using prompts like:
beautiful, masterpiece, best quality, perfect lighting,
For more 2.5D images use:
photorealistic, (hyperrealistic:1.2), beautiful, masterpiece, best quality, perfect lighting,
Mixed with the models:
FAQ
Comments (11)
Posted a comparison of Version 3 vs Version 1 which I was still on. 3 is better in some ways, but 1 has better coloring. 3 is more 2.5d while 1 is flatter
Both are definitely good, but are very different to the point of nearly being separate models from my tests
3 is better for some images for sure, but I feel like it's still worth keeping 1 around as well. Great model!
Thank you for the great feedback! I have actually been thinking about making seperate models now that v1 differs so much from v3. As you say, its as if v1 could be its own model.
I have been thinking about making something more stylized focused, something more semi realistic, and what i got now, something more 2.5D anime.
All models will be build up from the initial models i first started with, which would be dark sushi mix, rev animated and pikas animated mix. This is because i would like each model to have at least a hint of each other in them.
If you switch to V2 you will see many of my images. I think V2 is my favorite in regards to the bold colors and more anime like faces. V3 is great as well and I can see why people like it but for more mature (anime themed) faces I think V2 is better. This is my #1 favorite model for Anime creation with triple AAA screenshot quality.
Thank you for the comment. I'm working on making models with different themes in mind. Right now, I'm making a somewhat painting/stylized model which I will soon release. All models will have a merge of my V1 model which was my first model I posted
@BixBit11 I think your main image for V2 was the best advertising imo. It really portrayed the boldness and realistic/anime faces I like to render in. For that reason I like to use others loras mainly with your V2 to test. I hope to see new versions soon and will test always. V2 is chefs kiss.
Originally, the V2 was quite good, with great features. Now the V3 is kind of underwhelming. I like your model, come on, maybe the next version will be better。
can get unpruned version?
Bix any chance of getting this made into a SDXL version. I've been using this since you first put it out but now that SDXL is out i still want to use this but in that format, So, Whats the probability of this being made into XL ?
Tbh I have no idea how to convert the model to sdxl. And if I have to train it instead, it won't be possible since this is a merged model. I have not tried sdxl yet so for now, the probability is pretty low.
Thank you for the Pruned version! Much Love!
I have so much fun with this model. It dishes out great responses to prompts and offers huge variety. Really accurate, able to handle tons of loras and just truly an amazing work. Thank you!