schibsted/WAAS: Whisper as a Service
schibsted/WAAS is a powerful tool that offers a GUI and API for OpenAI Whisper with queuing functionality. It allows users to upload and transcribe audio or video files easily.
Key Features:
- The editor works 100% locally in the browser, enabling users to listen to segments and fix transcription errors.
- Users can directly download the transcribed results in various formats, including Jojo-file, SRT, or text, via email.
- It supports multiple models and allows users to customize settings such as language and task.
API Documentation: The API provides several endpoints for transcribing, detecting language, and downloading the results. Users can add a new transcribe job to the queue, get the available options for the transcribe and detect routes, and receive the finished job result in the requested output format.
Webhook Response: If a webhook ID is used in the request parameters, a POST will be sent to the specified webhook URL. The request contains a signature header for verification.
Contributing Requirements: The required amount of VRAM depends on the model used. The codebase is compatible with Python 3.8 - 3.10.
Installation and Running: Users can install and run the service using docker-compose or devcontainers. For GPU acceleration, specific configurations are required in the docker-compose.yml file.
In summary, schibsted/WAAS is a comprehensive and useful tool for audio and video transcription, offering a seamless experience for users and developers alike.
