Audio2Face-3D
此 Soniqo 页面记录本地 speech-swift / speech-core 实现中的 Audio2Face-3D。Hugging Face 包链接放在集成说明之后。
先进入站内页面
首页卡片和文档菜单先指向这里；源模型和权重包链接仍在本页提供。
概览
| 模型 | Audio2Face-3D |
|---|---|
| 用途 | Speech-to-avatar motion coefficients for downstream renderers |
| 后端 | MLX |
| 输出 | JSONL frames: timestamp plus the full coefficient vector |
| 语言 | Language-independent, audio-driven |
| 许可证 | NVIDIA Open Model License |
| 状态 | Ready through speech avatar-motion and the Audio2Face3D Swift product |
| 来源 | NVIDIA Audio2Face-3D |
| Swift 产品 | Audio2Face3D |
| CLI / 运行时 | speech avatar-motion |
使用
下面的片段对应当前 speech-swift 仓库暴露的 API 或命令。
# Generate avatar motion coefficients from speech audio.
.build/release/speech avatar-motion input.wav --output motion.jsonl --verbose
.build/release/speech avatar-motion input.wav \
--model aufklarer/Audio2Face-3D-v2.3.1-Claire-MLX模型链接
实现说明
- Three identities are published: James (default) and Claire at v2.3.1 with 169 coefficients, Mark at v2.3 with 301; skin coefficients live in the identity's own geometry basis, so renderers need the matching rig or a retarget projection.
- Two-branch regressor: a trimmed Wav2Vec2 audio encoder blended across transformer layers, plus a windowed-autocorrelation frequency branch, fused through emotion-conditioned strided convolutions.
- Each 8320-sample window of 16 kHz mono audio (hop 4160) emits one frame, conditioned on a 26-value emotion vector — 16 implicit plus 10 explicit values.
- The MLX graph runs the full NVIDIA model forward pass and is parity-checked against an ONNX fixture generated from the official release.