# Attribution

Open Swara voices are derived from reference samples sourced from the following open datasets. All voices were processed and enhanced using a proprietary humanization pipeline.

## Source Datasets

| Dataset | License | Voices | Link |
|---------|---------|--------|------|
| Piper TTS | MIT | 2,648 | [github.com/rhasspy/piper](https://github.com/rhasspy/piper) |
| Kyutai CC0 | CC0 1.0 | 228 | [github.com/kyutai-labs](https://github.com/kyutai-labs) |
| Mozilla Common Voice | CC0 1.0 | 200 | [commonvoice.mozilla.org](https://commonvoice.mozilla.org) |
| Bark | MIT | 131 | [github.com/suno-ai/bark](https://github.com/suno-ai/bark) |
| GLOBE V2 | CC0 1.0 | 100 | [kaggle.com/datasets/rtatman/speech-accent-archive](https://www.kaggle.com/datasets/rtatman/speech-accent-archive) |
| LibriTTS-R | CC-BY 4.0 | 79 | [openslr.org/141](https://www.openslr.org/141/) |
| Meta MMS-TTS | CC-BY-NC 4.0 | 38 | [huggingface.co/facebook/mms-tts](https://huggingface.co/facebook/mms-tts) |
| CosyVoice | Apache 2.0 | 15 | [github.com/FunAudioLLM/CosyVoice](https://github.com/FunAudioLLM/CosyVoice) |
| Qwen3-TTS | Apache 2.0 | 9 | [github.com/QwenLM/Qwen2.5-Omni](https://github.com/QwenLM/Qwen2.5-Omni) |
| Chatterbox Samples | MIT | 6 | [github.com/resemble-ai/chatterbox](https://github.com/resemble-ai/chatterbox) |

## License Notes

- **38 voices** from Meta MMS-TTS are licensed under CC-BY-NC 4.0 (non-commercial use only). These are in the following languages: Swahili, Telugu, Malayalam, Lao, Twi, and others from the MMS dataset.
- **3,416 voices** are fully commercial-use compatible.
- All other voices are under permissive licenses (MIT, CC0, Apache 2.0, CC-BY 4.0).

## How to Cite

```
@misc{openswara2026,
  title={Open Swara: The Largest Open-Source Humanized Voice Library},
  author={Jay Munshi},
  year={2026},
  url={https://github.com/jaymunshi/open-swara}
}
```