Recommended settings for this model (for version 1.0, 1.1 SemiReal):
In any case, you can set the settings that are more convenient for you.
Sampling steps from 35 to 50 (the faces are better and generally better)
Sampling method Euler a, DPM++2M Karras, DDIM,
Denoising strength 0.5 or 0.7
Be sure to tick Hires.fix
CFG Scale 7.5 or 8.5 (you can do more, you should play here)
Promt: ((masterpiece)), (high quality), (best quality), (detailed), hd, perfect lighting, detailed face, detailed body, (and then what you want)
Negative promt: (bad-hands-5:0.8),(easynegative:0.8), By bad artist -neg, negative_hand-neg, By bad artist, (worst quality, bad quality, child, cropped:1.4) ((monochrome)), ((grayscale)), sketches, text, speech bubble, artist name, signature, border, (bad_prompt_version2:0.8), (bad-artist-anime:0.8), duplicate, ugly, huge eyes, (bad proportions), bad perspective, multiple views, concept art, reference sheet, mutated hands and fingers, interlocked fingers, twisted fingers, excessively bent fingers, more than five fingers, lowres, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, (you can take something from this or leave your own, at your discretion)
Models 1.0 and 1.2 have been removed as they were a test version.
Description
FAQ
Comments (3)
Some images indeed come out with a subtle, but noticeable, figurine/doll effect. Feels a bit like those figurine/nendoroid LoRAs at 0.6 strength
Did you get rid of the original Anidoll model? The ones I have are Anidoll 1.0 and 2.0Semireal (I like them both differently). Is 1.1 Semireal an update on the 2.0 Semireal model?
Fantastic model, been running my usual line up of prompts to see the results and this has been the best model in months. Bravo.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.