Soniox STT V6 Adds
Soniox is a niche player next to Deepgram or AssemblyAI, but the trick here is genuinely neat: conditioning the recognizer on the agent's own transcript effectively replaces classic acoustic echo cancellation with text-level filtering. Self-looping voice bots remain the number-one complaint in real voice-agent deployments, so this targets a real and expensive pain point rather than a synthetic benchmark.… more
Expect larger STT vendors to fold similar context conditioning into their streaming APIs within a couple of product cycles. In our view, the deeper trend matters more than the release itself: duplexing logic keeps migrating from audio DSP into the model layer, and Soniox is just early on that curve.
Soniox STT v6 Adds Agent Text Input and Echo Filtering for Voice Agents
Soniox has released Soniox STT v6, the latest generation of its speech-to-text API. Soniox STT v6 is a speech-to-text model that reads what a voice agent says while hearing the caller. It uses that text to recognize short replies and to drop the agent’s own voice. The release …
Source:
Marktechpost
Discussion (0)