WAIT!
This model is useless... Unless used for close-up portraits or with FaceDetailer/inpainting. This model will not improve the eyes in a wide shot with distant eyes! I repeat: Only for close-up portraits, FaceDetailer, or inpainting.
So, when should you use the model?
During upscaling (improves, but isn't optimal)
Using FaceDetailer (works well)
With inpainting: manual or automatic segmentation of the eyes (best results)
Tips:
Works better by inpainting both eyes at once than a single one (risk of asymmetry)
The amount of denoising required will depend on how badly the eyes are drawn and the resolution of the inpaint mask:
Badly generated eyes and/or high resolution (> 1024 px) -> High denoise (0.4 to 0.6)
Acceptable eyes and/or lower resolutions -> Lower denoise (0.2 to 0.4)
Description
Just forget about the v1
FAQ
Comments (1)
The biggest difference between V2 and V1 is that, under the same conditions, the upper part of the eyes does not emit a dim light, or in other words, the level of darkness is lower.
Overall, both are good LoRA models.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.