Skip to content

CLI.md

Latest commit

 

History

History
87 lines (59 loc) · 4.16 KB

CLI.md

File metadata and controls

87 lines (59 loc) · 4.16 KB
Raw

AutoSubs Command Line Interface

AutoSubs can run from the terminal without opening a window, so AI agents and terminal-heavy users can transcribe files directly. Given a file argument the app runs headlessly, prints the result, and exits with a status code (0 success, 1 on a runtime error, 2 on a usage error). With no arguments it launches the normal desktop interface.

Basic Usage

# Readable transcript to the console (default format)
autosubs interview.mp4 --model small

# Speaker diarization (adds "Speaker N:" labels)
autosubs interview.mp4 --diarize --max-speakers 2 --lang en

# Pick a format explicitly…
autosubs interview.mp4 -f srt
autosubs interview.mp4 -f json

# …or let the output file extension decide
autosubs interview.mp4 -o subs.srt
autosubs interview.mp4 -o transcript.json

# Full option list / version
autosubs --help
autosubs --version

Model Selection

--model accepts any AutoSubs model — Whisper sizes (tiny…large-v3), parakeet, or a moonshine-* variant. Run autosubs --list-models for the full list.

Additional Options

Subtitle formatting:

  • --density - Text density: less, standard, more, single, or custom
  • --max-lines - Maximum lines per subtitle
  • --max-chars-per-line - Custom max characters per line (use with --density custom)
  • --text-case - Text case: none, lowercase, uppercase, or titlecase
  • --remove-punctuation - Strip punctuation from transcript

Translation:

  • --translate - Translate transcript to English
  • --target-language - Target language code for translation

Performance:

  • --gpu / --no-gpu - Force GPU on/off (default: auto-detect)

Advanced:

  • --prompt - Custom prompt to guide transcription

Output Formats

Format Contents
text (default) Readable transcript — [HH:MM:SS] Speaker N: …, one paragraph per speaker turn (no word-level timings)
srt SubRip subtitles (one short cue per segment)
vtt WebVTT subtitles (one short cue per segment)
json Full structured transcript including word-level timestamps

If --format is omitted, the format is inferred from the -o file extension (.srt, .vtt, .json, .txt), otherwise it defaults to text.

Output Behavior

stdout carries only the rendered output, so autosubs file.mp4 -f srt > out.srt is clean and pipe-safe. Progress and errors go to stderr: in an interactive terminal you get a live progress bar with the current stage (downloading model / transcribing / diarizing / translating); when stderr is piped or captured, it falls back to one line per stage. On failure a { "error": "..." } object is printed to stderr and the exit code is non-zero. Models are downloaded automatically on first use to the model cache.

On Windows, release builds attach to the parent console at startup so output is visible. As with any Tauri CLI app, the shell prompt may return before output finishes printing.

Getting the autosubs command on your PATH

The CLI is the same binary as the desktop app, so it needs to be reachable from your shell:

  • Linux — already done. The .deb/.rpm installs /usr/bin/autosubs, which is on PATH. Just run autosubs <file> ....
  • macOS / Windows — open Settings → Command line in the app and click Install. This symlinks the command into /usr/local/bin (macOS, prompts for your password) or adds the install folder to your user PATH (Windows). Remove reverses it. The button reports the current state on each platform.

During development the headless binary is at src-tauri/target/debug/autosubs (run cargo build in src-tauri first); symlink it yourself with ln -s "$(pwd)/src-tauri/target/debug/autosubs" /usr/local/bin/autosubs.

Exit Codes

  • 0 - Success
  • 1 - Runtime error
  • 2 - Usage error (invalid arguments)

Related Documentation