---
language:
- en
tags:
- face
- swap
- Fusion
- Model
- list
- Detection
- Alignment
- Enhancement
- Restoration
- Analysis
---
# FaceFusion 3.0.0 – Complete ONNX Model Repository

This repository contains the **full, unmodified ONNX model set** duplicated from  
`facefusion/models-3.0.0`.

All models are provided **for inference only** and are intended for use with **FaceFusion**
or custom ONNX Runtime pipelines.

Each `.onnx` file is paired with a `.hash` file used for **integrity verification**.

---

## General Information

- **Framework:** ONNX
- **Origin:** facefusion/models-3.0.0
- **Status:** Inference-ready
- **Execution:** CPU / CUDA / TensorRT / DirectML / CoreML (provider dependent)
- **Training:** Not supported

---

## Complete  List

### Landmark Detection & Alignment
- `2dfan4.onnx`
- `fan_68_5.onnx`

### ArcFace / Identity Conversion
- `arcface_w600k_r50.onnx`
- `arcface_converter_ghost.onnx`
- `arcface_converter_simswap.onnx`

### Face Detection
- `retinaface_10g.onnx`
- `scrfd_2.5g.onnx`
- `yoloface_8n.onnx`

### Face Segmentation
- `bisenet_resnet_34.onnx`
- `dfl_xseg.onnx`

### Face Swapping
- `inswapper_128.onnx`
- `inswapper_128_fp16.onnx`
- `simswap_256.onnx`
- `simswap_unofficial_512.onnx`
- `uniface_256.onnx`
- `blendswap_256.onnx`
- `ghost_1_256.onnx`
- `ghost_2_256.onnx`
- `ghost_3_256.onnx`

### Face Restoration & Enhancement
- `gfpgan_1.2.onnx`
- `gfpgan_1.3.onnx`
- `gfpgan_1.4.onnx`
- `gpen_bfr_256.onnx`
- `gpen_bfr_512.onnx`
- `gpen_bfr_1024.onnx`
- `gpen_bfr_2048.onnx`
- `restoreformer_plus_plus.onnx`
- `codeformer.onnx`

### Face Attributes / Analysis
- `fairface.onnx`
- `styleganex_age.onnx`

### Live Portrait / Motion Transfer
- `live_portrait_feature_extractor.onnx`
- `live_portrait_motion_extractor.onnx`
- `live_portrait_eye_retargeter.onnx`
- `live_portrait_lip_retargeter.onnx`
- `live_portrait_generator.onnx`
- `live_portrait_stitcher.onnx`

### Lip Synchronization
- `wav2lip_96.onnx`
- `wav2lip_gan_96.onnx`

### Super-Resolution & Upscaling
- `real_esrgan_x2.onnx`
- `real_esrgan_x2_fp16.onnx`
- `real_esrgan_x4.onnx`
- `real_esrgan_x4_fp16.onnx`
- `real_esrgan_x8.onnx`
- `real_esrgan_x8_fp16.onnx`
- `real_hatgan_x4.onnx`
- `ultra_sharp_x4.onnx`
- `lsdir_x4.onnx`
- `nomos8k_sc_x4.onnx`
- `span_kendata_x4.onnx`
- `clear_reality_x4.onnx`

### Colorization
- `deoldify.onnx`
- `deoldify_artistic.onnx`
- `deoldify_stable.onnx`
- `ddcolor.onnx`
- `ddcolor_artistic.onnx`

### Audio / Classification / Misc
- `kim_vocal_2.onnx`
- `open_nsfw.onnx`
- `peppa_wutz.onnx`

---

## Hash Files

Every model includes a corresponding `.hash` file:

- Ensures file integrity
- Required by FaceFusion
- Must not be removed or modified

---

## Example Usage

```python
import onnxruntime as ort

session = ort.InferenceSession(
    "inswapper_128.onnx",
    providers=["CUDAExecutionProvider", "CPUExecutionProvider"]
)