Recommended prompts: exposure blend, medium shot, bokeh, (hdr:1.4), high contrast, (cinematic, orange and white film), (muted colors, dim colors, soothing tones:1.3), low saturation, (hyperdetailed:1.2), (noir:0.4). You can apply these keywords to any image you want to generate, and you can modify the variable words to achieve the desired effect.
prompt：exposure blend, medium shot, bokeh, (hdr:1.4), high contrast, (cinematic, orange and white film), (muted colors, dim colors, soothing tones:1.3), low saturation, (hyperdetailed:1.2), (noir:0.4)
Recommended Parameters:
Please enable the Hires upscaler and use Afterdetail to redraw the face
Sampler: DPM++ 2M Karras or DPM++ SDE Karras
Steps: 20~40
Hires upscaler:4x-UltraSharp R-ESRGAN 4x+Anime6B
Hires upscale: 2
Hires steps: 15-40
Denoising strength: 0.2~0.6
CFG scale: 5-14
clip skip 2
Description
FAQ
Comments (5)
vamos!! dame 10 fotos de referencia por lo menos
Great model! However, it's not really FP16. I used the Model Toolkit extension on it and it shrunk the file down to 2 Gb by converting it to FP16
@Ilandis could you please upload an FP16 version?
FP32 or FP16 isn't necessarily what dictates a larger model file. Plenty of other things happen when running utilities such as Model Toolkit that can reduce it's binary file size.
Does converting from 32-bit floating point precision, down to 16-bit precision reduce size? Sure does. Does that mean you are correct in stating this isn't actually a 16-bit precision model? Maybe...
Just something to note though, so as to not further the proliferation of expert statements from experts on the internet.
Toolkit makes a Pruned version of the model