ruslantau/media-annotator

Web-based annotation tool for media data. The easiest way to create you own media dataset.

Media-annotator

Web-based annotation tool for media data.

Features

• Uploading selected audio-files from directory. Currently, only .wav .mp3 files are supported.
• Manual and auto transcribing for 20+ languages and dialects via Vosk

  ℹ️ Auto annotation will use the first channel of .mp3.

• Export of the speech regions as zip of wavs and regions.csv or regions.json

How to use

1. Run app as docker image

   docker-compose up -d

2. Open app page

How to debug

1. Install conda/miniconda/micromamba and node+npm

2. Clone repository, create python environment using conda manager and activate it

   git clone https://github.com/ruslantau/media-annotator
   cd annotator
   conda env create -f backend/environment.yaml
   conda activate annotator

3. Run FastAPI backend

   python backend/main.py

4. Install dependencies and run Nuxt frontend

   cd frontend
   npm install
   npm run build 
   npm run start

5. Open app page

TODO

• add docker images and setup CI/CD
• extend the list of supported formats (mp4,flac,avi,etc.)
• running auto annotation on selected region
• adding punctuation
• speaker diarisation