Product Overview: Whisper as a Service (WAAS)
Introduction
Whisper as a Service (WAAS) is a comprehensive solution developed by Schibsted, designed to provide an efficient and user-friendly interface for audio and video transcription using OpenAI’s Whisper model. This service combines a graphical user interface (GUI) with a robust API, making it an ideal tool for various transcription needs.
Key Features and Functionality
Transcription Capabilities
- Audio and Video Transcription: WAAS allows users to upload audio or video files and transcribe them into text. The transcription process is handled asynchronously by worker nodes, ensuring efficient processing.
User Interface (Jojo)
- GUI for Upload and Transcription: The Jojo GUI enables users to upload files, initiate transcription jobs, and receive download links for the transcribed content via email. Users can download files in various formats such as Jojo files, SRT (SubRip Subtitle), or plain text.
Editor
- Local Browser Editor: The editor works entirely within the user’s browser, allowing users to listen to segments of the audio/video and correct transcription errors. Saved files can be easily managed on the user’s desktop.
API
- Transcribe API: The API allows for adding new transcription jobs to the queue. It supports various parameters such as email callbacks, webhooks, language detection, and model selection. Users can also detect the language of the audio file using a dedicated API endpoint.
Job Management
- Job Status and Download: Users can check the status of their transcription jobs and download the results in multiple formats, including JSON, plain text, SRT, and WebVTT. This flexibility ensures that the transcribed content can be integrated into various applications.
Webhooks
- Webhook Integration: WAAS supports webhook notifications, allowing users to receive updates on job status. The webhook response includes a signature header for content verification, ensuring the integrity of the data.
Additional Features
- Automatic Language Detection: The service can automatically detect the language of the uploaded audio/video files.
- Model Selection: Users can choose from different Whisper models (e.g., ‘tiny’) to balance between accuracy and resource usage.
- Optional Parameters: Additional parameters such as task type (transcribe or translate) and filename customization are available to tailor the transcription process to specific needs.
Benefits
- Efficiency: WAAS streamlines the transcription process with asynchronous job processing and automated email notifications.
- Flexibility: The service offers multiple output formats and integration options through its API, making it versatile for different use cases.
- Accuracy: Leveraging OpenAI’s Whisper model, WAAS ensures high accuracy in transcription, enhanced by user-friendly correction tools.
Overall, Whisper as a Service (WAAS) is a powerful tool for anyone needing reliable and efficient audio and video transcription, with a user-friendly interface and robust API capabilities.