valle-models / README.md
qczone's picture
Initial model release (1.0.0)
cca0606
|
Raw
History Blame Contribute Delete
986 Bytes
---
tags:
- onnx
- coreml
---
# Valle Models
Model weights for local audio, image, and video processing.
| Directory | Model | Purpose |
| --- | --- | --- |
| `birefnet/` | BiRefNet | Foreground extraction and background removal |
| `demucs/` | HTDemucs | Separate vocals and instrumental audio |
| `dpdfnet/` | DPDFNet | Speech denoising and enhancement |
| `edgetam/` | EdgeTAM | Prompt-based video object segmentation and tracking |
| `lama/` | LaMa | Fill masked image regions and remove unwanted objects |
| `modnet/` | MODNet | Portrait matting and background removal |
| `omnishotcut/` | OmniShotCut | Detect shot boundaries and transition types |
| `realesrgan/` | Real-ESRGAN | 4x image upscaling |
| `rife/` | RIFE | Video frame interpolation |
| `transnetv2/` | TransNetV2 | Detect video shot boundaries |
Each model directory contains its weights, `release.v1.json`, and `card/LICENSE`.
`catalog.v1.json` indexes the available releases. Licenses are model-specific.