Articulate Armadillo: Audio to SRT Subtitle Generator

Generates SRT subtitles from MP3/WAV audio using Whisper or Faster-Whisper backends. Optimized for developer use.

Core Features

• Backends & Models: --backend whisper (OpenAI) or faster-whisper. Supports various model sizes (e.g., tiny to large-v3).
• Performance: --device cpu or cuda. --test mode for 3-min audio clips (auto-creates if non-existent).
• Output Quality: Segments split to sentences, proportional timing, 80-char/2-line SRT formatting.
• Efficiency: Caches models, skips redundant test clip creation. faster-whisper shows duration-based progress.

Setup

1. Prerequisites: Python 3.x, FFmpeg (in PATH or update main.py).
2. Environment: python -m venv venv && source venv/bin/activate (or venv\Scripts\activate on Windows).
3. Dependencies: pip install -r requirements.txt.

Usage

1. Audio files (MP3/WAV) in input/.
2. python main.py [OPTIONS]
3. Subtitles in output/.

CLI Options

Argument	Choices	Default	Description
--test			Use 3-min test clip.
--backend	whisper, faster-whisper	whisper	Transcription backend.
--model	(backend-dependent)	large	Model size (e.g., tiny, large, large-v3).
--device	cpu, cuda	cpu	Execution device.

Refer to backend docs for specific model availability.

Notes

• For full dependency list, see requirements.txt.
• Handles Whisper model checksum errors by offering cache clearing.
• Ensure CUDA environment is correctly configured if using --device cuda.