Hear2Read Promises Machine-learned Audio Translations for India

By Gary Cutlack on at

Text to speech tools are nothing new, but the one offered by Hear2Read -- for free -- is a little bit different from the rest. It uses machine learning to populate a database and compare the commonalities of Indian languages, offering translation on the fly without a massive bespoke list of words.

Developed in tandem with the Carnegie Mellon University, Hear2Read was created with the main ambition of making something to work on low-spec devices that could be distributed for free. And there it is, on Google Play for Android, promising to translate documents into spoken words for people speaking Tamil across India and around the world.

The system adapts a baseline structure after recording two or three hours of raw speech, with a machine learning tool apparently able to adapt its output depending on the dialect. The problem is that India has 22 official languages, so future versions of the tool will need to work pretty hard to cater for everyone.

Alan Black, from the university's Language Technologies Institute, explained: "Each language is different and historically TTS systems have been done one at a time. We looked at commonalities of Indian languages and developed tools to apply the same technology to multiple languages." [Hear2Read via TechCrunch]

Want more updates from Gizmodo UK? Make sure to check out our @GizmodoUK Twitter feed, and our Facebook page.