jianchang512/stt
Stt
This project is a hardware-accelerated transcription server and offline subtitle generator. It functions as a speech-to-text tool that converts audio and video files into plain text, JSON, and SRT subtitle formats using the Whisper model.
The system operates as an OpenAI Audio API emulator, providing a local server that mimics a specific audio interface. This allows it to serve transcriptions to existing client configurations without requiring changes to the client software.
The service utilizes GPU acceleration to increase voice recognition speed and includes utilities for hardware detection and resource consumption optimization. It supports offline model inference for data privacy and provides post-processing capabilities to convert text between simplified and traditional scripts.
Features
- Speech to Text Transcription - Transcribes audio and video files into text, JSON, or subtitles using offline recognition models.
- OpenAI-Compatible Endpoints - Provides API endpoints that mimic the OpenAI specification to support third-party client integrations.
- Transcription APIs - Provides a network endpoint for uploading media files to receive automated text transcriptions.
- Speech Recognition Engines - Provides an optimized inference engine for converting audio to text using Transformer-based models locally.
- Speech-to-Text Transcribers - Provides a tool for transcribing audio and video into text using the Whisper model.
- Subtitle Generators - Generates SRT and JSON subtitle files with timing information from audio and video recordings.
- Offline Transcript Processors - Processes audio and video locally using pre-trained models without requiring an internet connection.
- GPU Hardware Acceleration - Leverages GPU hardware acceleration to speed up the mathematical computations required for voice recognition.
- Local Media Processing - Processes voice recordings into text on local hardware to ensure data privacy.
- OpenAI-Compatible Audio Servers - Implements a local HTTP server that mimics the OpenAI Audio API for seamless integration with existing clients.
- Accelerated Transcriptions - Uses optimized GPU compute engines to significantly increase the speed of speech-to-text transcription.
- Multi-Format Output Converters - Converts internal transcription data into multiple formats including plain text, JSON, and SRT subtitles.
