Subtitling Tools

The award-winning subtitling software was developed by the British Broadcasting Corporation and is now available under an open-source licence.

White Paper: Speech Recognition in Assisted and Live Subtitling for Television

Software components

Assisted Subtitling

Assisted Subtitling (AS) uses document analysis and speech recognition to produce a set of subtitles (closed captions) for a TV programme or other audio-visual content. A recording of the programme is required, together with a transcript.

Dialogue is extracted from the transcript, and tagged with information regarding which character says it, and in which scene. Speech recognition then aligns the extracted dialogue with the programme's soundtrack, establishing the start and end time for each spoken. Word timings, togther with the timings of shot changes in the video are then processed by an expert system to make an optimal set of subtitles. These subtitles are coloured, to allow characters to be easily distinguished, formed to give high readability, and are well-timed with respect to shot changes and the soundtrack.

kLive

This is for live material where a subtitler re-speaks or re-keys the dialogue in real time. Client software is provided for the subtitlers to use and server software is provided to receive data from the clients and send subtitle data to the appropriate inserter. A monitoring client is also provided.

Speech User Guide

Steno User Guide

Reversioning

A web server tool for processing subtitle files to make them conform more closely to BBC style. Typical use is to reformat subtitles on programmes originating from North America.

Software

Project page

The source code can be accessed via anonymous CVS or can be browsed online directly from the CVS repository.

Contact

The preferred way to contact the developers of Subtitling Tools is via the SourceForge forums.

The Subtitling Tools software was written by Michael Evans, John Fletcher and Matthew Marks. Feedback is welcome.