audio-transcriber

What problem does it solve?

This Skill converts dashcam video audio into accurate, timestamped transcripts and optionally assigns speaker labels, enabling quick review and searchable archives of driving conversations.

Core Features & Use Cases

• GPU-accelerated transcription: Uses faster-whisper with CUDA to transcribe large dashcam video collections quickly.
• Speaker diarization (optional): Produces unified speaker labels (spkA, spkB, etc.) across transcripts for multi-speaker sessions.
• Multi-format outputs: Generates TXT, JSON, SRT, and optional VTT transcripts, plus a global INDEX.csv for search across days.
• Date/folder organization: Outputs organized by date and video, with a central results JSON and per-day indices for cross-day analysis.
• Use Case: Imagine a fleet manager needing to review driver conversations across hundreds of drives; this Skill batches the work, tags who spoke when, and provides easily searchable transcripts.

Quick Start

1. Place your dashcam videos in a date-stamped directory (e.g., MYY/MM/DD or YYYYMMDD within a base folder).
2. Run the skill to transcribe a folder, enable optional diarization, and export TXT/JSON/SRT (and INDEX.csv) to a chosen outputs directory.
3. Inspect the produced INDEX.csv to search phrases across all transcripts and jump to exact time stamps.