Transcriptor - create transcripts from News Video

Status: active
How might we support journalists, archivists and technical staff with accurate transcription services in News?

BBC News produces hundreds of hours of raw footage and edited material every day, much of which is made up of interviews. Finding and communicating the relevant parts of them is a huge challenge and journalists are frequently asked to transcribe material for a variety of uses.

Transcriptor is a transcription tool using the Kaldi “Speech to text” engine in Window on the Newsroom. It takes audio input from BBC production systems and pushes it through Kaldi. Transcriptions are generated and made available to journalists, technical staff and archivists.

The aim is to allow journalists to check and correct the tool’s work, rather than type everything themselves - so freeing time for more original journalism.

If you want to find out more, send us a tweet @bbc_news_labs

Next Priorities

  1. Transcriptor viewer - clickable transcripts, time alignment of the source audio/video and transcript section
  2. Transcriptor editor - Potential integration with radio R&D prototypes
  3. Inline speaker ID and tags against the timeline