Basic Usage

Command Syntax

macscribe <INPUT> [--model MODEL] [--output OUTPUT]

Arguments

<INPUT> (required): URL or local file path - YouTube: https://youtube.com/watch?v=... - Apple Podcasts: https://podcasts.apple.com/... - X/Twitter: https://x.com/username/status/... - Local file: ./audio.mp3 or /path/to/video.mp4

--model (optional): Whisper model - Default: mlx-community/whisper-large-v3-mlx - Example: --model mlx-community/whisper-medium-mlx

--output / -o (optional): Save to file - File path: transcript.txt - Directory: ./transcripts/ (auto-generates filename) - Creates parent directories automatically

Supported Formats

Audio: MP3, WAV, FLAC, M4A, OGG, WMA Video: MP4, MOV, AVI, MKV, WEBM, M4V, WMV

macscribe audio.mp3
macscribe video.mp4

Output

Transcripts are automatically copied to clipboard. Use Cmd+V to paste.

Save to file:

# Specific file
macscribe audio.mp3 --output transcript.txt

# Directory (auto-generates filename)
macscribe audio.mp3 --output ./transcripts/

How It Works

URL sources: Downloads audio → Transcribes → Copies to clipboard → Cleans up temp files

Local files: Validates file → Extracts audio (if video) → Transcribes → Copies to clipboard

Model Performance

Model	Accuracy	Speed	RAM
Large (default)	Excellent	~5-10x real-time	~8GB
Medium	Very Good	~3-5x real-time	~4GB
Small	Good	~2-3x real-time	~2GB
Base	Fair	~1-2x real-time	~1GB

macscribe audio.mp3 --model mlx-community/whisper-medium-mlx

Requirements

Apple Silicon (M1/M2/M3+) for MLX acceleration
RAM varies by model (2-8GB)