Description
ScribeCrate Live is a robust, open-source solution designed for private, real-time speech-to-text transcription and speech-to-English translation. It allows users to deploy and operate a transcription server on their own hardware, ensuring data privacy and control. The system is powered by WhisperLive and faster-whisper, leveraging advanced Whisper models for high accuracy and efficiency.
Read More
The core functionality revolves around processing live microphone audio and various audio streams, converting them into progressive text segments as words are spoken. This real-time capability is delivered over WebSocket, making it suitable for applications that require immediate text output, such as live captioning or interactive voice interfaces. For speech-to-English translation, users can select a suitable multilingual Whisper model and configure the task accordingly.
Beyond live streaming, ScribeCrate Live also provides an OpenAI-compatible REST API for transcribing uploaded audio files. This dual approach caters to different input flows, offering flexibility for developers integrating transcription services into their applications. The server runs locally, with fresh persistent installations generating and storing API keys for secure authentication, ensuring that transcription processes remain private and under user control.
Deployment is streamlined through Docker, supporting both CPU (amd64 and arm64 architectures) and NVIDIA GPU acceleration with a dedicated CUDA image. Users can choose from various Whisper models, with larger models offering higher accuracy but requiring more memory and processing power. Voice Activity Detection (VAD) is included to intelligently skip silent periods, optimizing resource usage. For offline operation, models can be pre-cached, and the system can be configured to run without internet access for model downloads.
ScribeCrate Live is part of a larger ecosystem, including the Self-Hosted AI Stack, allowing users to integrate it with other local AI tools for chat, document processing, and more. The project is open source under the MIT License, meaning users are responsible for their own hardware or cloud hosting costs. It emphasizes user control over data and processing, making it a strong choice for privacy-conscious developers and organizations.
- CategoryAI Transcription Tools
- Pricing ModelOpen Source
- DeploymentSelf-Hosted
- PlatformsDocker (CPU, NVIDIA GPU)
ScribeCrate Live's Core Features
Real-time speech-to-text transcription
Speech-to-English translation
WebSocket streaming for live audio
OpenAI-compatible REST API for file transcription
Self-hosted deployment for privacy and control
Supports various Whisper models
Voice Activity Detection (VAD) to skip silence
CPU (amd64, arm64) and NVIDIA GPU (CUDA) support
Docker deployment for easy setup
Persistent model cache for offline operation
API key authentication for secure access
Open-source under MIT License
Integrates with Self-Hosted AI Stack
Progressive text output as words arrive
Client setup documentation available on GitHub
How to use ScribeCrate Live?
Prepare Host: Ensure a supported Linux host with Docker and sufficient memory for your chosen Whisper model.
Start Server: Deploy ScribeCrate Live using Docker, connecting live audio or testing with a recording.
Retrieve API Key: For persistent installations, retrieve the generated API key for authenticated access.
Connect Client: Use a compatible WebSocket client for live microphone input or the REST API for file transcription.
Configure Task: Set `translate=True` in the Python client or `"task": "translate"` in WebSocket handshake for speech-to-English translation.
Select Model: Choose a suitable multilingual Whisper model (e.g., base, small, medium, large-v3) for translation tasks.
Test Transcription: Replace `audio.mp3` with your file and run the provided REST request for an authenticated test.
ScribeCrate Live's Use Cases
- Live Captioning
- Microphone Transcription
- Speech-to-English Translation
- Audio File Transcription
- Building Voice Applications
- Private Transcription Services
- Self-Hosted AI Stack Integration
