What problem does it solve?
Converting existing recordings, podcasts, meetings, or videos into text usually requires uploading files to cloud services or manually configuring Whisper environments. This Skill transcribes local audio and video files entirely on-device, producing ready-to-use Markdown transcripts and SRT subtitle files.
Core Features & Use Cases
- Local Whisper Transcription: Runs mlx-whisper on Apple Silicon (Apple GPU via MLX) or faster-whisper on CPU/CUDA machines, with automatic engine and model selection.
- Markdown and SRT Output: Generates only
.md transcripts and .srt subtitles, with language detection, duration, and model metadata included.
- Batch and Long-Recording Handling: Supports multiple files, output directories, VAD silence filtering, and repetition-loop prevention for long Chinese recordings.
- Use Case: A user drops a two-hour Chinese meeting recording (
meeting.m4a) and asks for the text; the Skill transcribes it locally with whisper-large-v3-turbo-q4 and returns the transcript plus saved Markdown and SRT files.
Quick Start
Use the local-audio-transcriber skill to transcribe my recording.m4a file into Chinese text with Markdown and SRT output.
