transcribe-cli

A modular audio transcription CLI with pluggable providers. Agent-friendly.

Currently supports AssemblyAI with speaker diarization, sentiment analysis, entity detection, auto chapters, summarization, topic detection, and more.

Installation

# With uv (recommended)
uv tool install transcriber-cli

# With pip
pip install transcriber-cli

# Run without installing
uvx --from transcriber-cli transcribe audio.mp3

Or install from source:

git clone https://github.com/miguelarios/transcribe-cli.git
cd transcribe-cli
uv tool install .

Setup

Get an API key from the AssemblyAI Dashboard and set it:

export ASSEMBLYAI_API_KEY='your_key_here'

Or pass it directly: transcribe --api-key <key> audio.mp3

Usage

# Basic transcription (text output to stdout)
transcribe interview.mp3

# JSON output
transcribe interview.mp3 -f json

# With speaker diarization
transcribe meeting.wav --speaker-labels

# SRT subtitles to file
transcribe lecture.mp3 -f srt -o lecture.srt

# Full-featured transcription
transcribe call.m4a --speaker-labels --entities --sentiment -f json

# Pipe JSON to jq
transcribe interview.mp3 -f json | jq '.segments[] | .speaker, .text'

# Agent-friendly summary (saves full transcript to temp file)
transcribe meeting.mp3 --summary

# Transcribe from URL
transcribe https://example.com/audio.mp3

Output Formats

All Options

Run transcribe -h for the full list. Key options:

Speaker/Diarization

• --speaker-labels — Enable speaker diarization
• --speakers-expected N — Exact speaker count
• --min-speakers N --max-speakers M — Speaker range
• --speaker-id-type [role|name] — Speaker identification mode
• --speaker-names NAME — Known speaker names (repeatable)

Analysis

• --entities — Detect names, locations, dates, etc.
• --sentiment — Sentiment analysis per utterance
• --topics — IAB topic detection
• --auto-chapters — Generate chapters with headlines
• --summarize — Generate transcript summary
• --content-safety — Content moderation

Language

• --language CODE — Language code (e.g., en_us)
• --language-detection — Auto-detect language

Advanced

• --prompt TEXT — Context prompt for transcription
• --keyterms TERM — Domain-specific terms to boost (repeatable)
• --multichannel — Multichannel transcription
• --redact-pii — Redact personally identifiable information
• --filter-profanity — Filter profanity
• --disfluencies — Include filler words (um, uh)

Output

• -f, --format [text|json|srt|vtt] — Output format
• -o, --output PATH — Write to file instead of stdout
• --summary — Output metadata summary, save full transcript to temp file
• -v, --verbose — Show progress and timing on stderr

Meta

• --list-providers — List available transcription providers
• --api-key KEY — AssemblyAI API key
• -V, --version — Show version
• -h, --help — Show help

Development

git clone https://github.com/miguelarios/transcribe-cli.git
cd transcribe-cli
uv run --extra dev pytest tests/ -v

License

MIT