Skip to main content

Transcribe videos to ASS subtitle files in 99 languages using faster-whisper

Project description

whispersub

Transcribe video files to ASS subtitle files using faster-whisper. Whisper detects the spoken language automatically and supports 99 languages, including English, Spanish, French, German, Japanese, Chinese, Arabic, Hindi, and many more.

  • NVIDIA GPU acceleration with automatic CPU fallback
  • Batch mode: pass multiple video files or directory trees
  • Surround-sound audio extraction (dialogue-channel aware)
  • Word-level timestamps with balanced line breaking
  • Per-word confidence colour coding in the terminal

Requirements

Python 3.10+. Works on Linux, Windows, and macOS. No system FFmpeg needed — PyAV bundles its own. Supports MKV, MP4, AVI, MOV, WebM, TS, and other common video formats.

Install

pip install whispersub

GPU acceleration (Linux/Windows — requires an NVIDIA GPU with CUDA 12):

pip install whispersub[gpu]

Without [gpu], whispersub falls back to CPU automatically if CUDA is unavailable.

On first run, whispersub downloads the Whisper large-v3-turbo model (~800 MB) from Hugging Face and caches it locally.

Usage

# Single file — writes movie.en.ass alongside the video
whispersub movie.mkv

# Explicit output directory
whispersub movie.mkv --output-dir ~/subs

# Whole directory, force overwrite
whispersub /media/shows --force

# File with multiple audio tracks — inspect, then pick one
whispersub series.mkv --list-audio-tracks
whispersub series.mkv --audio-track 2

Options

Option Default Description
--audio-track N Audio track index (required if the file has multiple tracks)
--list-audio-tracks off Show audio tracks for all input videos, grouped by configuration, and exit
--colour-by probability Per-word terminal colour coding: probability or duration
--font-size N 48 Font size (1280×720 canvas; player scales to actual resolution)
--force off Overwrite existing subtitle files (keeps backups)
--keep N 3 Number of .bak copies to keep when overwriting
--limit N Stop after N segments per video (useful for testing)
--max-line-count N 2 Maximum subtitle lines per card
--max-line-width N 36 Maximum characters per line
--max-threads N all cores CPU thread limit
--output-dir DIR alongside video Write all subtitle files to this directory

Output

Subtitle files are named <stem>.<language>.ass, e.g. movie.en.ass. The detected language comes from Whisper. Output is compatible with VLC, mpv, IINA, MPC-HC, and other players that support ASS/SSA subtitles.

We chose ASS over SRT for better-looking subtitles: font sizing scales correctly to any resolution, and line breaks are balanced for readability. It also allows us to preserve word-level timing, so the file can be post-processed or reformatted without re-transcribing.

Licence

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

whispersub-1.5.0.tar.gz (33.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

whispersub-1.5.0-py3-none-any.whl (17.9 kB view details)

Uploaded Python 3

File details

Details for the file whispersub-1.5.0.tar.gz.

File metadata

  • Download URL: whispersub-1.5.0.tar.gz
  • Upload date:
  • Size: 33.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for whispersub-1.5.0.tar.gz
Algorithm Hash digest
SHA256 280ba8aaab293ded87fdf71c572cf0d9aa952b79e69c66db10af5c951719582a
MD5 285877d96e09cd4850d85cc22c744aa7
BLAKE2b-256 229957cbd4bd8ac9ea5e43991be55a7345b1cca24f0acee5f304e566761dbc9d

See more details on using hashes here.

Provenance

The following attestation bundles were made for whispersub-1.5.0.tar.gz:

Publisher: publish.yml on zvea/whispersub

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file whispersub-1.5.0-py3-none-any.whl.

File metadata

  • Download URL: whispersub-1.5.0-py3-none-any.whl
  • Upload date:
  • Size: 17.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for whispersub-1.5.0-py3-none-any.whl
Algorithm Hash digest
SHA256 d834ccf366cb648174d651a4b789dad199f291f01551c6a70d0c3da9b0cb8212
MD5 8db95b64a561282bbecb7ce35a6aae09
BLAKE2b-256 d31a21bebf7beed63fd9c453a7f60e7a0ef4cb8dd8c97dd37e37db923a1fa4b8

See more details on using hashes here.

Provenance

The following attestation bundles were made for whispersub-1.5.0-py3-none-any.whl:

Publisher: publish.yml on zvea/whispersub

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page