---
title: audiogen-medium-mlx
canonical_url: "https://www.modelscope.cn/models/mlx-community/audiogen-medium-mlx"
md_url: "https://www.modelscope.cn/models/mlx-community/audiogen-medium-mlx.md"
repository: mlx-community/audiogen-medium-mlx
last_updated: 2026-05-29
license: cc-by-nc-4.0
model_type:
  - audiogen
base_model:
  - facebook/audiogen-medium
base_model_relation: finetune
parameters: 2.2B
tensor_type:
  - F16
  - F32
library_name:
  - mlx
  - safetensors
  - pytorch
frameworks:
  - pytorch
downloads: 48
stars: 0
tags:
  - audio-generation
  - text-to-audio
  - audiogen
  - mlx
  - encodec
---

# audiogen-medium-mlx

> audiogen-medium-mlx - mlx-community 在 ModelScope 开源的模型。AudioGen Medium (MLX)

mlx-community/audiogen-medium-mlx 是 ModelScope 魔搭社区上的 2.2B 参数机器学习模型，采用 cc-by-nc-4.0 许可，基于 facebook/audiogen-medium 构建。

- **Repository**: mlx-community/audiogen-medium-mlx
- **License**: cc-by-nc-4.0
- **Parameters**: 2.2B
- **Base model**: facebook/audiogen-medium
- **Tags**: audio-generation, text-to-audio, audiogen, mlx, encodec
- **Downloads**: 48
- **Stars**: 0
- **Last updated**: 2026-05-29

Source: https://www.modelscope.cn/models/mlx-community/audiogen-medium-mlx

---

# AudioGen Medium (MLX)

This is the MLX-native port of [facebook/audiogen-medium](https://huggingface.co/facebook/audiogen-medium), a 1.5B parameter autoregressive transformer for text-to-audio generation.

## Model Details

- **Architecture**: Autoregressive Transformer LM over EnCodec discrete tokens
- **Parameters**: ~1.5B (LM) + EnCodec compression model
- **Sampling rate**: 16 kHz
- **Frame rate**: 50 Hz (4 codebooks, delayed pattern)
- **Text encoder**: T5-large (d_model=1024, 24 layers, 16 heads)
- **Max duration**: 10 seconds (configurable)

## Files

- `config.json` — Model configuration (includes `t5_model_name` reference)
- `model.safetensors` — LM + EnCodec weights
- `model.safetensors.index.json` — Weight index (for sharded variants)

### T5 Conditioner (extracted separately)

The T5-large text encoder weights are not included in this repository. Use `extract_t5.py` to extract them from the original `facebook/audiogen-medium` checkpoint:

```bash
python extract_t5.py --output /path/to/audiogen-mlx/t5
```

This produces a `t5/` directory with `config.json`, `model.safetensors`, and tokenizer files.

> **Note**: The T5 safetensors keys use MLX-compatible naming (`.layer_0.` / `.layer_1.`
> instead of HuggingFace's `.layer.0.` / `.layer.1.`). This is required because MLX's
> `ModuleParameters.unflattened()` splits on all dots.

## Usage (Swift/MLX)

```swift
import MLXAudioGen

let model = try await AudioGenModel.fromPretrained(
    modelFolder: modelURL,
    t5Folder: t5URL
)

let tokens = try await model.generate(
    descriptions: ["dog barking"],
    duration: 5.0,
    cfgCoef: 3.0,
    temperature: 1.0,
    topK: 250
)

let audio = model.decode(tokens: tokens)
```

## T5 Attention

T5's self-attention intentionally does **not** scale scores by `1/sqrt(d_k)`. This is a deliberate design choice in the T5 architecture — do not add scaling in the inference code.

## License

This model is published under the [CC-BY-NC 4.0](https://creativecommons.org/licenses/by-nc/4.0/) license (non-commercial use only), following the original [AudioGen license](https://huggingface.co/facebook/audiogen-medium).