lyriks-video 0.3.0

Automated lyrics video generator

Lyriks

Lyriks is an automated lyrics video generator. It transcribes the audio and automatically creates a video using MoviePy.

---

Features

• Automatic vocal separation using Demucs
• Transcription with OpenAI Whisper and whisper-timestamped
• Synchronized lyrics video generation with MoviePy

---

Requirements

• Linux
• An NVIDIA GPU (recommended for best performance; CPU is supported but slower)
• 10GB of free disk space
• Python 3.11

---

Installation

It is highly recommended to use a virtual environment for isolation:

python3 -m venv .venv
source .venv/bin/activate

Then install Lyriks with pip:

pip install lyriks-video

---

Usage

python -m lyriks generate AUDIO_FILE LYRICS_FILE [OPTIONS]

Parameters

• AUDIO_FILE
  Path to the input audio file (e.g., song.mp3).
  This should be a supported audio format (such as MP3 or WAV).

• LYRICS_FILE
  Path to the lyrics file (plain text).
  The lyrics should be in a text file, one line per lyric segment.

Options

You will be interactively prompted in the CLI for any options you leave unspecified.

• --output, -o
  Output video file name (without extension).
  Example: -o my_lyrics_video

• --model_size, -m
  Sets the Whisper model size for transcription.
  Options: tiny, base, small, medium, large, turbo

• --device, -d
  Which device to use for Whisper model inference.
  Options: cpu, cuda

• --generator, -g
  Which backend to use for video generation.
  Options:

  • mp: MoviePy (slow, low quality)
  • ps2: pysubs2 + ffmpeg (fast, good quality, experimental)
  • ts: Only save transcript (for debugging)

• --no-gemini
  Disable Gemini improvements for Whisper output.

---

Example

python -m lyriks generate path/to/song.mp3 path/to/lyrics.txt -m small -d cuda -o output_video

Note: This process can take up to 20 minutes on lower end hardware.

---

TODO

• Fix up lyrics using Gemini
• Per-word highlighting in videos
• Fancier video styles and effects
• Add more robust error handling
• Ask which background to use
• Batch processing

---

Credits

This project uses Demucs for music vocal separation.

@inproceedings{rouard2022hybrid,
  title={Hybrid Transformers for Music Source Separation},
  author={Rouard, Simon and Massa, Francisco and D{'e}fossez, Alexandre},
  booktitle={ICASSP 23},
  year={2023}
}

@inproceedings{defossez2021hybrid,
  title={Hybrid Spectrogram and Waveform Source Separation},
  author={D{'e}fossez, Alexandre},
  booktitle={Proceedings of the ISMIR 2021 Workshop on Music Source Separation},
  year={2021}
}