stt

To transcribe audio messages into text using a GPU, this Skill leverages the faster-whisper small model in a dedicated Python virtual environment with CUDA acceleration. It automatically processes voice messages when invoked via transcription keywords.

Faster-whisper works without a CUDA-enabled GPU by automatically falling back to CPU processing for speech-to-text. This ensures transcription remains functional even when GPU hardware is unavailable, though processing speed will be reduced.

To set up the environment for faster-whisper speech-to-text, configure CUDA, create a dedicated Python virtual environment, and install required dependencies. Environment and model configuration details are documented in env_config.md for quick setup and replication.

The best way to perform audio transcription for Chinese voice messages is using the faster-whisper small model with GPU acceleration. It provides fast and reliable Chinese speech-to-text transcription, reducing manual effort significantly.

Yes, you can use this speech-to-text model for languages other than Chinese. The faster-whisper small model supports GPU-accelerated audio transcription for Chinese and other languages, providing reliable text output across multiple language contexts.

To invoke the speech-to-text transcription process, call whisper_stt.py with the path to an audio file. Alternatively, the transcription triggers automatically when users send voice messages or use keywords like 'speech-to-text' or 'transcription'.