|
|
--- |
|
|
license: other |
|
|
license_name: stability-ai-community |
|
|
license_link: https://huggingface.co/stabilityai/stable-audio-open-small/blob/main/LICENSE |
|
|
extra_gated_heading: "Access to Stability AI Models Required" |
|
|
extra_gated_description: "This model requires access to Stability AI's original model. Please request access to stable-audio-open-small first, then come back here." |
|
|
extra_gated_button_content: "I have Stability AI access" |
|
|
extra_gated_prompt: "By requesting access, you confirm that you have been granted access to Stability AI's stable-audio-open-small model on Hugging Face." |
|
|
extra_gated_fields: |
|
|
Stability AI Username: text |
|
|
I confirm I have access to stable-audio-open-small: checkbox |
|
|
--- |
|
|
|
|
|
# Stable Audio Open Small - TensorFlow Lite + Runner |
|
|
|
|
|
Complete package for running Stability AI's `stable-audio-open-small` model on edge devices. |
|
|
|
|
|
**Powered by Stability AI** |
|
|
|
|
|
## Contents |
|
|
|
|
|
- `conditioners_float32.tflite` - T5 text encoder model |
|
|
- `dit_model.tflite` - Diffusion transformer model |
|
|
- `autoencoder_model.tflite` - Audio decoder model |
|
|
- `spiece.model` - SentencePiece tokenizer model |
|
|
- `audiogen.exe` - Inference runner binary |
|
|
|
|
|
## Quick Start |
|
|
|
|
|
```bash |
|
|
./audiogen.exe ./ "ambient forest sounds" 4 42 |
|
|
``` |
|
|
|
|
|
Arguments: `<models_path> <prompt> <num_threads> <seed>` |
|
|
|
|
|
The runner will generate an `output.wav` file with 10 seconds of stereo audio at 44.1kHz. |
|
|
|
|
|
## Model Description |
|
|
|
|
|
This is a converted version of the original [stable-audio-open-small](https://huggingface.co/stabilityai/stable-audio-open-small) model by Stability AI, split into three optimized TensorFlow Lite models for mobile and edge deployment. Includes a ready-to-use C++ inference runner. |
|
|
|
|
|
**Architecture:** |
|
|
- **Text Encoder**: T5-based conditioner for prompt processing |
|
|
- **Diffusion Model**: DiT (Diffusion Transformer) for latent generation |
|
|
- **Audio Decoder**: Autoencoder for converting latents to audio |
|
|
|
|
|
## Components & Licenses |
|
|
|
|
|
| Component | License | Copyright | |
|
|
|-----------|---------|-----------| |
|
|
| **AI Models** (*.tflite) | Stability AI Community License | Stability AI Ltd. | |
|
|
| **Tokenizer** (spiece.model) | Apache 2.0 | Google | |
|
|
| **Runner Binary** (audiogen.exe) | Apache 2.0 | ARM Limited | |
|
|
|
|
|
## Usage Rights |
|
|
|
|
|
- ✅ **Research & Non-Commercial**: Free to use |
|
|
- ✅ **Commercial Use**: Free for organizations with <$1M annual revenue (registration required at [stability.ai](https://stability.ai/community-license)) |
|
|
- ⚠️ **Commercial Use**: License terminates if annual revenue exceeds $1M (enterprise license required) |
|
|
|
|
|
## License |
|
|
|
|
|
This repository is primarily governed by the **Stability AI Community License Agreement** due to the AI models. |
|
|
|
|
|
The SentencePiece tokenizer and ARM inference code components are under Apache 2.0, but the overall package follows the most restrictive license (Stability AI Community License). |
|
|
|
|
|
See the complete license terms: [Stability AI Community License](https://huggingface.co/stabilityai/stable-audio-open-small/blob/main/LICENSE) |
|
|
|
|
|
## Attribution |
|
|
|
|
|
This Stability AI Model is licensed under the Stability AI Community License, Copyright © Stability AI Ltd. All Rights Reserved. |
|
|
|
|
|
**Original Model**: [stable-audio-open-small](https://huggingface.co/stabilityai/stable-audio-open-small) by Stability AI |
|
|
**Conversion**: TensorFlow Lite conversion by InnerMost47 |
|
|
**Runner**: Based on ARM's TensorFlow Lite inference code |
|
|
|
|
|
## Disclaimer |
|
|
|
|
|
This package is provided "AS IS" without warranties. The TensorFlow Lite conversion may affect model performance. Please refer to the original model documentation for detailed specifications. |
