ReFocus Model Dependencies
This repository serves as a backup mirror of the model dependencies required by the ReFocus project.
All models originate from the lllyasviel repositories (misc and fooocus_inpaint) and are used under their respective open-source licenses. u2net.onnx is sourced from the danielgatis/rembg repository.
License & Attribution
- All models in this repository are redistributed under the terms of the GNU General Public License v3.0.
- This is a mirror for backup and availability purposes only.
- ReFocus and this repository do not claim ownership of these models.
- Original models are created and maintained by lllyasviel, danielgatis, and other respective authors.
Included Models
From lllyasviel/misc (14 files)
| Model File | Purpose |
|---|---|
clip_vision_vit_h.safetensors |
CLIP Vision model for IP-Adapter |
control-lora-canny-rank128.safetensors |
Canny ControlNet LoRA |
fooocus_ip_negative.safetensors |
Negative embedding for IP-Adapter |
fooocus_upscaler_s409985e5.bin |
Upscaler model (ESRGAN) |
fooocus_xl_cpds_128.safetensors |
CPDS ControlNet LoRA |
ip-adapter-plus-face_sdxl_vit-h.bin |
Face IP-Adapter for SDXL |
ip-adapter-plus_sdxl_vit-h.bin |
IP-Adapter for SDXL |
model_base_caption_capfilt_large.pth |
BLIP captioning model (Describe feature) |
sdxl_lcm_lora.safetensors |
LCM LoRA for SDXL |
vaeapp_sd15.pth |
VAE approx for SD 1.5 |
wd-v1-4-moat-tagger-v2.csv |
WD14 tagger label mapping |
wd-v1-4-moat-tagger-v2.onnx |
WD14 tagger ONNX model (Anime Describe) |
xl-to-v1_interposer-v3.1.safetensors |
VAE interposer for refiner swap |
xlvaeapp.pth |
VAE approx for SDXL |
From lllyasviel/fooocus_inpaint (4 files)
| Model File | Purpose |
|---|---|
fooocus_inpaint_head.pth |
Inpainting head |
inpaint.fooocus.patch |
Inpainting patch (v1) |
inpaint_v25.fooocus.patch |
Inpainting patch (v2.5) |
inpaint_v26.fooocus.patch |
Inpainting patch (v2.6) |
From danielgatis/rembg (1 file)
| Model File | Purpose |
|---|---|
u2net.onnx |
Rembg U²-Net for mask generation |
Note:
u2net.onnxis not part oflllyasvielrepositories. It is sourced from the official rembg release. See danielgatis/rembg for more information.
From facexlib (2 files)
| Model File | Purpose |
|---|---|
parsing_parsenet.pth |
Face parsing model for face detection/cropping |
detection_Resnet50_Final.pth |
Face detection model for face cropping |
Note: These files are sourced from the facexlib repository and are used for face-related features (e.g., FaceSwap IP-Adapter preprocessing).
Original Source
- ReFocus: https://github.com/OliverDestiny/ReFocus
- Defooocus: https://github.com/ehristoforu/DeFooocus
- Fooocus: https://github.com/lllyasviel/Fooocus
- Original Models: https://huggingface.co/lllyasviel
- Rembg (u2net): https://github.com/danielgatis/rembg
- facexlib: https://github.com/xinntao/facexlib
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support