Agent skill · ma08
transcribe
Transcribe audio/video files using speech-to-text providers. Use when user has audio or video files to transcribe.
What it needs
About 2k tokens when loaded.
What this skill does
Transcribe Skill Transcribe audio and video files to text using speech-to-text providers (currently Soniox). When to Use User has audio or video files to transcribe User wants to convert meeting recordings, interviews, or media to text User needs batch transcription of a directory of files User mentions transcription, speech-to-text, or converting audio/video to text How to Use Single file Directory of files With domain context and terms Arguments Reference Flag Short Description Default ------ ------- ------------- --------- --input -i Audio/video file or directory (required) -- --output-dir -o Output directory Same as input --provider -p STT provider (soniox) soniox --context -c Free-text context for accuracy "" --terms -t Comma-separated domain terms "" --language -l Language hint ISO code en --no-cleanup Keep remote files after transcription false --no-combined Skip combined transcript for directories false Supported Formats .mov, .mp4, .m4a, .mp3, .wav, .webm, .ogg, .flac, .aac, .aiff, .amr, .asf Output Files For each input file example.mp4: example-transcript.json - Raw API response with tokens example-transcript.md - Readable markdown with speaker labels For multi-file runs (unless --no-combined): combined-transcript.md - All transcripts in one file Requirements SONIOXAPIKEY must be set in the environment or in ~/pro/personalos/.env Keep the machine-local .env mode 0600; never copy the key into task artifacts or source control. Use the canonical ~/pro/personalos/scripts/transcribe.py path so the command works regardless of the current directory. Get an API key at https://console.soniox.com Typical Workflow 1. User provides audio/video file(s) 2. Run the transcribe script with appropriate context/terms 3. Review the generated markdown transcript 4. Use the transcript for meeting notes, documentation, or content Examples in Context Meeting recording: Batch factory tour videos:
How to use it
Reference it in AdaL, Claude Code, Cursor or any coding agent — nothing to install:
@skills ma08/transcribe