Attributions
Last updated: 2026-06-12. Audexum is built on open-source software and open-weight AI models. This page lists the third-party components whose licences require attribution or whose terms are material to how the Service operates.
AI models
| Model / component | Licence | Notes |
|---|---|---|
| NVIDIA Parakeet-TDT-0.6B-V2 (speech recognition) HuggingFace | CC-BY-4.0 | Copyright © NVIDIA Corporation. This model is used under the Creative Commons Attribution 4.0 International licence. Attribution is required by the licence terms. |
| Supertonic-3 (default TTS engine) HuggingFace | OpenRAIL-M | BigScience Open RAIL-M licence. Use restrictions (Attachment A) flow down to end users — see our Acceptable Use Policy § 0a. |
| Kokoro ("Audexum HD" English voices) HuggingFace | Apache-2.0 | Kokoro is used as a secondary TTS engine for English voices. |
| AudioSeal (AI watermarking, Meta) GitHub | MIT | Copyright © Meta Platforms, Inc. Being integrated to embed invisible AI-generation markers in Output, as required by Article 50 of the EU AI Act. |
Key open-source components
| Package | Licence | Use in Audexum |
|---|---|---|
| FastAPI | MIT | REST API framework |
| Uvicorn | BSD-3-Clause | ASGI server |
| yt-dlp | Unlicense (public domain) | YouTube transcript extraction tool |
| ONNX Runtime | MIT | Model inference runtime (GPU) |
| React | MIT | Frontend framework |
| Vite | MIT | Frontend build tool |
| Tailwind CSS | MIT | UI styling |
Full dependency lists are maintained in requirements.txt (Python backend) and package.json (frontend). All packages are used under their respective open-source licences.
Contact
If you believe a required attribution is missing or incorrect, please contact [email protected].