PolyAI Releases Dialog-RSN-1: An Audio-Native Dialog Model That Fuses Turn-Taking, Speech Recognition, Function Calling, And Response
PolyAI's Dialog-RSN-1 fuses speech recognition, turn-taking, and function calling into one audio-native model — sub-300ms latency in production.
Why it matters
A new model architecture that handles end-to-end voice dialog without ASR transcription step, with production latency below the perceptibility threshold. This is a genuine capability advance for voice agents, and practitioners building voice products should track whether this architecture becomes table-stakes.
The key facts
7 to knowAudio-native model architecture (raw audio → response, skipping ASR intermediate)
Fuses: turn-taking detection, speech recognition, function calling, response generation
Sub-300ms response latency reported in live deployments
TTS kept separate for voice control/customization
Request-based (not always-on streaming) execution model
Model name: Dialog-RSN-1
Vendor: PolyAI
Go to the source
MarkTechPostmarktechpost.com
Publisher excerpt: PolyAI has introduced Dialog-RSN-1, a dialog model that perceives caller audio directly instead of reading an ASR transcript. It fuses turn-taking, speech recognition, function calling, and response generation into a single audio-native model, keeps TTS separate so the output voice stays…