ATTRIBUTION.md

Attributions

voice-server

This project incorporates open-source components and was inspired by existing work in the text-to-speech community.

Direct Dependencies

• MLX-audio by MLX Community

  • License: MIT
  • Repository: https://github.com/ml-explore/mlx-audio
  • Used for: Fast local TTS with Kokoro-82M model (Apple Silicon)

• Kokoro-82M by LMSTUDIO

  • License: Apache 2.0
  • Repository: https://huggingface.co/mlx-community/Kokoro-82M-bf16
  • Used for: Text-to-speech model with 41 built-in voices

• Bun by Jarred Sumner

  • License: MIT
  • Repository: https://github.com/oven-sh/bun
  • Used for: TypeScript runtime and server framework

Historical References (Previously Used)

The following components were used in earlier versions but have been removed:

• Qwen3-TTS by Alibaba Qwen Team (Apache 2.0) - Previously used for custom voice cloning
• FastAPI by Sebastián Ramírez (MIT) - Previously used for Python web server
• pyttsx3 by Natesh MB (LGPLv3) - Previously used for macOS TTS fallback

Architectural Inspiration

ValyrianTech/Qwen3-TTS_server

• Repository: https://github.com/ValyrianTech/Qwen3-TTS_server
• License: Apache 2.0
• Relationship: Historical architectural inspiration only

Our initial implementation was inspired by the ValyrianTech Qwen3-TTS_server reference implementation for:

• FastAPI server structure patterns
• TTS inference workflow design
• Endpoint organization concepts

However, our current implementation is original code with significant differences:

• MLX-audio backend instead of Qwen TTS
• TypeScript-only architecture (no Python subprocess)
• Kokoro-82M with 41 built-in voices
• ElevenLabs-compatible API for PAI integration
• Simplified single-backend architecture

No code was copied from ValyrianTech/Qwen3-TTS_server.

Related Projects

PAI Voice Server (Original)

• Repository: https://github.com/danielmiessler/Personal_AI_Infrastructure
• Our server is designed as a drop-in replacement for the ElevenLabs-based PAI voice server, maintaining API compatibility while using local TTS.

---

License Notice

This project (voice-server) is licensed under the MIT License. See LICENSE file for details.

Third-party components are licensed under their respective terms as listed above.