| name | ml-models |
| description | ML model tools: AudioCraft (MusicGen text-to-music, AudioGen text-to-sound) and Segment Anything (SAM zero-shot image segmentation). |
| version | 1.0.0 |
| author | Hermes Agent |
| license | MIT |
| platforms | ["linux","macos"] |
| metadata | {"hermes":{"tags":["ML","AudioCraft","MusicGen","SAM","image-segmentation","text-to-music"]}} |
ML Models Suite
Unified skill for specialized ML models: audio generation and image segmentation.
Section A: AudioCraft (MusicGen + AudioGen)
Generate music and sound effects from text prompts using Meta's AudioCraft.
MusicGen: Text-to-music generation. Create songs, instrumentals, and musical pieces from text descriptions.
AudioGen: Text-to-sound generation. Create sound effects, ambient sounds, and Foley from text.
Key features:
- Text prompt → audio generation
- Multiple model sizes (small, medium, large)
- Conditioning on melody/chords (MusicGen)
- Duration control
Typical workflow:
- Define text prompt describing the audio
- Select model size based on quality needs
- Generate audio → save as WAV/MP3
- Optionally post-process (trim, fade, normalize)
Section B: Segment Anything Model (SAM)
Zero-shot image segmentation via points, boxes, or masks using Meta's SAM.
Best for: Object segmentation without training, interactive segmentation, instance segmentation.
Key features:
- Zero-shot: no training data needed
- Multiple input modes: points, boxes, masks
- High-quality segmentation masks
- Works with any image
Typical workflow:
- Load image
- Provide segmentation input (point click, bounding box, or text prompt)
- SAM generates mask(s)
- Apply mask for downstream tasks (editing, analysis, extraction)
