Skip to main content

Speech Recognition

Status​

Production

This project is live and actively maintained.

Description​

Our recognition API.

Team​

Owner & Tech Lead

Location​

  1. Synchronous audio file recognition hosted on phoneticsrv3 (https://phoneticsrv3.lcs.tcd.ie/asr_api/test)
  2. Streaming recognition is accessible via websockets on the recognition-testing VM on Hetzner (wss://recognition-testing-api.abair.ie)
  3. Fotheidil recognition (plus speaker diarisation, capitalisation and punctuation) is hosted on Banba (proxied through the recognition VM on hetzner and also phoneticsrv3). Media files are first sent to the Fotheidil VM on Hetzner for processing, which then calls the recognition API (http://10.0.0.8:6060/generate_transcripts)

Example API Calls​

  1. See docs
  2. See new website websockets connection
  3. See Fotheidil API call