speech-align-viz 0.1.1

Audio Text Alignment Visualization Tool

SpeechAlignViz 🎙️📝

SpeechAlignViz is a web-based visualization tool for audio-text alignment. (🤗 Hugging Face) It allows researchers and developers to verify and interact with speech transcription segments (SRT, VTT, TextGrid, JSON) directly on the audio waveform.

Features

• Unified Waveform View: View text segments directly on the audio timeline.
• Multi-Format Support:
  • JSON: custom format [{text, start_time, end_time}]
  • SRT: SubRip Subtitle files
  • WebVTT: Web Video Text Tracks
  • TextGrid: Praat annotation files
• Interactive Control:
  • Click any segment to play its corresponding audio.
  • Zoom in/out for precise alignment checks.
  • Automatic audio-text synchronization.
• Local & Upload Support: Load files from your local filesystem or upload them directly via the browser.

Installation

From PyPI

pip install speech-align-viz

From Source

git clone https://github.com/LoganLiu66/SpeechAlignViz.git
cd SpeechAlignViz
pip install -e .

Usage

Start the server using the CLI:

speech-align-viz serve --port 8001 --host 0.0.0.0

Then open http://0.0.0.0:8001 in your browser.

CLI Figure Export

You can generate a static alignment figure directly from the command line:

speech-align-viz export \
  --audio ./examples/example.wav \
  --subtitle ./examples/example.json \
  --output-fig ./examples/example.png

Note: The figure width is automatically calculated based on audio length to prevent overcrowding, but you can override it with --width N.

Loading Data

1. Audio: Upload an audio file (mp3, wav, ogg) or provide a local path.
2. Transcript: Upload a support transcript file (.json, .srt, .vtt, .TextGrid) or provide a local path.

Development

The project structure is split into:

• src/speech_align_viz: FastAPI backend and CLI.
• frontend: React + Vite + WaveSurfer.js frontend.

To build the frontend:

cd frontend
npm run build

This will bundle the frontend into src/speech_align_viz/static.

License

MIT