Fluent learns in all accents and languages, including those that have no associated text corpora. It also lets people switch languages or combine several in the same sentence.
Our patented technology goes back to how children learn to talk, through examples in context rather than through text. This gives our speech recognition engine the flexibility to understand natural phrases reliably across languages, accents, and noise environments, whether that be a smart home, connected car, wearable or virtual reality headset.
Learning from users in their specific acoustic environments allows Fluent to achieve 95-100% accuracy across a variety of noise conditions, after only a few example phrases.
People should not have to learn hard-coded phrases from user manuals. We allow each user in a home or car to speak naturally and create fun voice shortcuts on the fly, without textual input.
Fluent works offline, sidestepping connectivity and privacy issues. It has a small footprint suitable for embedded systems. The cloud is accessed for updates or to recognize infrequent phrases.
Speech recognition’s rank in complaints from new car owners in the US
major languages each spoken by more than 70 M people (vs. 17 languages supported by Siri)
people native in a language not in the top 38
Meet the Team
Scientist with 7 years of speech recognition research experience. Formerly at Nuance. PhD in speech recognition from McGill University.
Formerly a strategy consultant at Roland Berger. BA from Harvard, PhD from Cambridge.
Developer formerly at RG/A and Zoobe. MSc in music signal processing from Queen Mary University of London.