ALTO - a multilingual journalism tool

Status: active
ALTO is a virtual voice-over tool for reversioning video content into multiple languages using text-to-speech voice synthesis
feature-image

ALTO is a virtual voice-over tool for reversioning video content into multiple languages using text-to-speech voice synthesis. It allows a single news editor to generate translated audio tracks with multiple voices.

ALTO combines a number of cutting edge language technologies, including computer assisted translation and text-to-speech voice synthesis. A video transcript is first generated using machine translation before it becomes available for manual editing. In the second step, language journalists convert the translated script into a computer-generated voice track using text-to-speech technology (TTS). ALTO then automatically attaches the new translated audio to the video file, on top of the natural sound track.

We launched a pilot service in Japanese in December 2015 and a second pilot in Russian in April 2016. We’ve since added new features, such as a pronunciation dictionary, synthetic speech mark-up language components, automated video trimming and name strap translation. The user now gets more control over the phonetic performance of the synthetic voices: they can modify the spelling of a word to trigger the correct pronunciation and save it into the dictionary. SSML (synthetic speech mark-up language) allows users to insert pauses to improve intelligibility of a sentence.

During our pilots we received positive and encouraging feedback from our audiences. About 80% expressed support of the synthetic voice technology. While machine translation and synthetic voice technology are by no means perfect, we have found that they are good enough to allow non-English speakers watch BBC news reports in their own language.

Background

ALTO is dedicated to assisting multilingual journalists in the translation and reversioning of video content. This not only includes translating the video script and generating a new audio track, but also — at a later stage – experimenting with automated subtitles generation and the translation of graphics and name straps.

More from around the Web


Next Priorities

  1. Iterate and improve the tools