Open Source Licenses & Acknowledgments
Brass Scripts is built on the shoulders of open source giants. We gratefully acknowledge the following projects and their contributors.
AI Models & Core Technologies
OpenAI Whisper
Our transcription service is powered by OpenAI's Whisper, a general-purpose speech recognition model trained on diverse audio data capable of multilingual speech recognition, speech translation, and language identification.
- Project: OpenAI Whisper
- License: MIT License
- Usage: Base speech recognition model for audio transcription
WhisperX
We use WhisperX for enhanced transcription with precise word-level timestamps and improved performance through faster-whisper and wav2vec2 alignment integration.
- Project: WhisperX
- License: BSD-4-Clause License
- Usage: Enhanced speech recognition with word-level timestamps
- Authors: Max Bain and others
pyannote.audio
Speaker diarization (identifying different speakers) is handled by pyannote.audio, an open-source toolkit offering state-of-the-art pretrained models and pipelines based on PyTorch.
- Project: pyannote.audio
- License: MIT License
- Usage: Speaker diarization and identification
- Authors: Hervé Bredin and contributors
faster-whisper
For improved performance, we utilize faster-whisper, a reimplementation of OpenAI's Whisper model using CTranslate2, offering significantly faster transcription with lower memory usage.
- Project: faster-whisper
- License: MIT License
- Usage: Optimized Whisper inference engine
- Authors: SYSTRAN and contributors
Supporting Libraries
PyTorch
The machine learning computations are powered by PyTorch, an open-source machine learning library providing tensor computation with GPU acceleration and deep neural networks.
- Project: PyTorch
- License: BSD-3-Clause License
- Usage: Machine learning framework for model inference
Transformers (Hugging Face)
Model loading and inference utilities are provided by Hugging Face's Transformers library, offering state-of-the-art natural language processing models and tools.
- Project: Transformers
- License: Apache 2.0 License
- Usage: Model loading and preprocessing utilities
Web Platform Technologies
Next.js
Our web application is built with Next.js, a React framework for production-grade applications with features like server-side rendering, API routes, and optimized performance.
- Project: Next.js
- License: MIT License
- Usage: Web application framework
React
The user interface is built with React, a JavaScript library for building interactive user interfaces with component-based architecture.
- Project: React
- License: MIT License
- Usage: Frontend user interface library
Tailwind CSS
Styling and responsive design are implemented using Tailwind CSS, a utility-first CSS framework for rapidly building custom user interfaces.
- Project: Tailwind CSS
- License: MIT License
- Usage: CSS framework for styling and layout
Infrastructure & Services
Supabase
Database services and real-time functionality are provided by Supabase, an open-source Firebase alternative built on PostgreSQL.
- Project: Supabase
- License: Apache 2.0 License
- Usage: Database and backend services
Node.js
Server-side processing and API endpoints run on Node.js, a JavaScript runtime built on Chrome's V8 engine for building fast and scalable network applications.
- Project: Node.js
- License: MIT License
- Usage: Server-side JavaScript runtime
License Compliance
We are committed to respecting and complying with all open source licenses. All dependencies are used in accordance with their respective licenses, and we extend our gratitude to the maintainers and contributors of these projects.
License Summary
- MIT License: Whisper, WhisperX, faster-whisper, pyannote.audio, Next.js, React, Tailwind CSS, Node.js
- BSD Licenses: WhisperX (BSD-4-Clause), PyTorch (BSD-3-Clause)
- Apache 2.0 License: Transformers, Supabase
Attribution & Credits
Special recognition goes to the researchers, developers, and maintainers who have made these technologies available to the open source community:
- OpenAI team for developing and open-sourcing the Whisper speech recognition model
- Max Bain and collaborators for WhisperX enhancements and word-level timestamp alignment
- Hervé Bredin and the pyannote.audio team for state-of-the-art speaker diarization
- SYSTRAN for the optimized faster-whisper implementation
- Hugging Face for the Transformers library and model ecosystem
- Meta (Facebook) for the PyTorch machine learning framework
- Vercel for Next.js and React ecosystem contributions
- Supabase team for the open-source backend platform
- All open source contributors who have improved these projects
Research Papers & Citations
Our service is built on academic research. If you use our transcripts in academic work, please consider citing the underlying research:
Whisper
Radford, A., Kim, J. W., Xu, T., Brockman, G., McLeavey, C., & Sutskever, I. (2022). Robust Speech Recognition via Large-Scale Weak Supervision. arXiv preprint arXiv:2212.04356.
WhisperX
Bain, M., Huh, J., Han, T., & Zisserman, A. (2022). WhisperX: Time-Accurate Speech Transcription of Long-Form Audio. arXiv preprint arXiv:2303.00747.
pyannote.audio
Bredin, H., & Laurent, A. (2021). End-to-end speaker segmentation for overlap-aware resegmentation. Proc. Interspeech 2021, 3111-3115.