● DEVOTION — A MULTILINGUAL FULL-DUPLEX VOICE API
Devotion is a full-duplex voice model you call over an API, priced like any other usage-based AI API. It hears, thinks and speaks at once, in every language — built on the best open duplex models and NVIDIA's open weights.
THESIS
Voice AI today is a walkie-talkie. It waits for you to finish, transcribes, thinks, then performs a reply.
Human conversation doesn't take turns. It overlaps — interruptions, backchannels, silences that mean something.
Full-duplex models finally live inside that timing. Devotion is the API that puts it in every language, for anyone building on top of it.
The duplex breakthrough already happened — PersonaPlex, Moshi and their kin listen and speak in one continuous stream. Rebuilding that would burn compute on a solved problem. Devotion starts from open weights and spends every GPU-hour on what's missing, then ships it as a single API endpoint.
Timing is cultural. How long a polite pause lasts, when overlap is warmth and when it's rudeness, what a backchannel sounds like — all of it differs by language. A multilingual duplex model has to learn each language's rhythm, not just its words.
Multilingual conversational speech, synthetic overlapping dialogue, and post-training that adds languages without breaking the duplex behaviours already there — with yield latency and backchannel timing benchmarked per language.
Half the economy runs on conversation — most of it not in English. Full-duplex makes machine voice employable; multilingual makes it global.
Agents that can be interrupted, corrected and talked over — and still hold the thread, in the customer's own language.
NEAR-TERMListening and speaking at once is the job description. Translation that keeps pace with the speaker — not after them.
NEAR-TERMReal-time conversational support for people who speak, hear or process differently. Timing is dignity: no dead air, no being talked over.
PILOTIntake and triage that listens the way a nurse does — probing, confirming, reassuring while the patient is still talking.
PILOTRobots, vehicles and machines that coordinate with people by voice, in places where hands and eyes are already busy.
RESEARCHThe Devotion API is opening to design and compute partners first. Join the waitlist to get access when it does.
Devotion is sold the way any usage-based AI API is sold: no seats, no minimums, pay for what you stream.
Billed on streamed audio-minutes, both directions. No plan, no lock-in — usage stops, billing stops.
Same endpoint, same price, regardless of language — no surcharge for going multilingual.
Published rate card lands at general availability. Waitlist members get early access and locked-in early-adopter pricing.
Ship the Devotion API. Streaming inference and infrastructure at conversational latency.
OPENINGPost-train Devotion on multilingual conversational data. Own model quality end to end.
OPENINGNo forms, no cover letters. Send the most interesting thing you've built to hello@instituteofthinking.com
JOIN THE WAITLIST
Early sign-ups get first access and locked-in early-adopter pricing when the API opens.