WAAS

WAAS: Whisper as a Service - GUI and API for OpenAI Whisper

WAAS (Whisper as a Service) is an open-source project that provides a GUI and API for OpenAI's Whisper, making audio and video transcription more accessible and user-friendly. It offers both a graphical user interface (GUI) for easy file upload and transcription and an API for programmatic access.

What is WAAS?

WAAS provides an interface to upload and transcribe audio or video files. After transcription, users receive an email with download links for the transcription in various formats, including Jojo-file, SRT, or plain text. A key feature is the local browser-based editor for correcting transcription errors.

Key Features

• GUI for Upload and Transcription: Simple interface for uploading audio and video files.
• Email Notifications: Receive email notifications with download links after transcription.
• Multiple Output Formats: Download transcriptions in Jojo-file, SRT, or plain text formats.
• Local Browser-Based Editor: Correct transcription errors within the browser.
• API Access: Programmatic access to transcription services via API.

How does WAAS work?

WAAS allows users to upload audio or video files through a GUI (named Jojo) or via an API. The uploaded file is then processed using OpenAI's Whisper model for transcription. Once the transcription is complete, the user receives an email containing links to download the transcription in various formats. The browser-based editor allows users to refine and correct any errors in the transcription before saving the final result.

API Documentation

The WAAS API provides several endpoints for transcription and related tasks:

• POST /v1/transcribe: Adds a new transcription job to the queue.
  • Required parameters: email_callback or webhook_id.
  • Optional parameters: language, model, task, filename.
  • Body: Raw audio data.
• OPTIONS /v1/transcribe: Retrieves available options for the transcription route.
• POST /v1/detect: Detects the language of the audio file.
  • Optional parameter: model.
  • Body: Raw audio data.
• OPTIONS /v1/detect: Retrieves available options for the detect route.
• GET /v1/download/<job_id>: Retrieves the completed transcription in the requested output format.
  • Optional parameter: output (json, timecode_txt, txt, vtt, srt).
• OPTIONS /v1/download/<job_id>: Retrieves available options for the download route.
• GET /v1/jobs/<job_id>: Retrieves the status and metadata of the specified job.
• GET /v1/queue: Retrieves the current length of the queue.

Webhook Integration

WAAS supports webhook notifications. Upon successful or failed transcription, a POST request is sent to the configured webhook URL with a JSON payload and an X-WAAS-Signature header for content verification.

Who is WAAS for?

• Researchers needing to transcribe interviews or lectures.
• Journalists working with audio or video content.
• Developers integrating transcription services into their applications.
• Anyone needing to quickly and accurately transcribe audio or video files.

Installation

To install and run WAAS, follow these steps:

1. Clone the repository.
2. Create a virtual environment.
3. Install the required Python packages using pip install -r requirements.txt.
4. Configure environment variables such as BASE_URL, EMAIL_SENDER_ADDRESS, EMAIL_SENDER_PASSWORD, and EMAIL_SENDER_HOST.
5. Run the setup using Docker Compose.

Running with Docker Compose

1. Create a .envrc file with the necessary environment variables.
2. Add a allowed_webhooks.json file (if using webhooks) with valid webhook URLs and tokens.
3. Run docker-compose --env-file .envrc up.

Using NVIDIA CUDA

To enable GPU acceleration with NVIDIA CUDA:

1. Install NVIDIA Docker.
2. Edit the docker-compose.yml file to use the Dockerfile.gpu and uncomment the device reservation.
3. Run docker-compose --env-file .envrc up.

Why choose WAAS?

WAAS offers a user-friendly interface and API for leveraging OpenAI's Whisper model. Its features like email notifications, multiple output formats, and local browser-based editing make it a convenient and efficient solution for audio and video transcription needs. The flexibility to run it locally or integrate it into existing systems via the API makes it a versatile tool for various use cases.

In conclusion, WAAS is a valuable tool for anyone looking to transcribe audio or video content quickly and accurately. Its open-source nature and ease of use make it an excellent choice for both personal and professional use.