Voice AI research lab
Voice AI for the languages people speak at home. Searchable text, captions and automated actions, built on our own research. Voices soon.
Who it's for
Every call in writing, in the language the caller chose.
Each call comes back timed to every word, with the language detected and any harmful speech flagged.
Train your own intents. Act on every call.
Teach Bhala what matters to your business: a payment question, a complaint, a cancellation. Every call and message is labelled with your intents, so your systems can route, alert or follow up on their own.
Broadcasters
Captions on air, about two seconds behind, in the language of the show.
Platforms and media
Subtitles for every video, machine or reviewed by a person, and harmful speech caught in voice notes and audio posts.
Voices that speak these languages.
Natural speech in isiZulu, isiXhosa, Sepedi and Kiswahili. Join the early list to use it first.
An estimated 3.4 billion people speak languages today's AI barely hears.
So we start with the language
Its sounds, its tones, how its words are built. Then we design speech models around them.
Then put it to work
Recordings, calls and broadcasts become timed text, captions and insight, in one place.
And open it to builders
The models behind Studio, through one API, in your own product.
Hear what your audience is saying.
60 free minutes to start, then $0.02 a minute.