# Meta AudioGen review (2026)

> Open sound-effect generation model (non-commercial weights).

AudioGen is part of Meta's AudioCraft research library and generates sound effects from text. The model weights are licensed for non-commercial use only.

**Best for:** Research and prototyping, not shipped games.

**Make AI Games score:** 54/100 (facts-based; hands-on test pending)

| Fact | Value |
|---|---|
| Website | https://github.com/facebookresearch/audiocraft |
| Status | Research demo |
| Categories | Sound effects |
| Pricing | Free tier; Free |
| Commercial use | Not allowed (open source) |
| Engines | Unverified |
| Exports | WAV |
| Skill level | advanced |
| Last verified | 2026-10-21 |

## Score breakdown

- Accessibility: 74/100
- Output & integration: 44/100
- Commercial rights: 10/100
- Value: 100/100
- Maturity: 20/100
- Game focus: 100/100
- Features: 65/100

## Features

- Text to sound effect: Yes (audiogen-medium 1.5B text-to-sound; needs 16GB GPU)
- Multiple variations: Partly (Python API generates a batch of samples per call with random sampling; no app UI)
- API: Partly (Python library you run yourself; no hosted web API)

## Pricing plans

- Open source: free (Open source, self-hosted; code MIT, weights CC-BY-NC 4.0)

License note: Code MIT; model weights CC-BY-NC 4.0 (non-commercial).

## Sources

- https://github.com/facebookresearch/audiocraft
- https://github.com/facebookresearch/audiocraft/blob/main/model_cards/AUDIOGEN_MODEL_CARD.md
- https://huggingface.co/spaces/facebook/MusicGen/raw/5325fcc37771a4fe5fc34a557e23b4988718602d/README.md
- https://huggingface.co/spaces/CryptoThaler/melody/blob/main/docs/AUDIOGEN.md
- https://github.com/facebookresearch/audiocraft/blob/main/docs/AUDIOGEN.md

Canonical: https://makeaigames.com/tools/meta-audiogen