# Miscellaneous Features Models

Models for high-performance video/audio generation miscellaneous features.

## Features Supported

### Audio
- **ACE-Step**: AI music generation from prompts + lyrics
- **Podcast**: Multi-speaker podcast generation (SoulX-Podcast-1.7B)

### Video Processing
- **Face Enhancer**: Video face enhancement/swap (ReActor + CodeFormer/GPEN)
- **RIFE Upscaler**: Video frame interpolation + upscaling
- **Lucy Edit**: Text-guided video editing
- **Fusion I2V**: 4-image to video generation (Phantom)
- **Animate Posenet**: Character animation with pose control (WAN 2.2 Animate)
- **Camera Control**: Camera movement video generation
- **Animate Photo**: Motion transfer from video to image

## Model Categories

| Category | Description |
|----------|-------------|
| `diffusion_models/` | Main diffusion models (Lucy Edit, Phantom, Animate, Camera) |
| `text_encoders/` | UMT5-XXL text encoders |
| `vae/` | WAN VAE decoders |
| `clip_vision/` | CLIP Vision encoder |
| `loras/` | Lightning and LightX2V LoRAs |
| `facerestore_models/` | Face restoration models (CodeFormer, GPEN) |
| `insightface/` | Face detection and swap models |
| `upscale_models/` | RealESRGAN upscaler |
| `rife/` | RIFE frame interpolation |
| `sams/` | SAM2 segmentation |
| `onnx/` | ONNX detection models (VitPose, YOLOv10) |
| `ace_step/` | ACE-Step music generation models |
| `TTS/SoulX-Podcast-1.7B/` | Podcast generation models |

## Usage with RunPod

Set in RunPod endpoint settings:
- **Model**: `Aloukik21/unreal-misc`
- **HF Token**: Your HuggingFace token

The models will be automatically cached and used by the misc Docker container.