Soniox releases STT v6 with agent text input and echo filtering

Soniox has released STT v6, a new speech-to-text model that adds agent text input and echo filtering. The update lets developers pass an agent’s text to the model while the voice agent is speaking, so the system can detect short user replies and filter out the agent’s own echo. The model also includes speaker diarization and language identification. Soniox said v6 is more accurate than v5, though it did not publish benchmark or performance figures. The release comes with two API options: the streaming endpoint stt-rt-v6 and the asynchronous endpoint stt-async-v6. According to the announcement, the model supports more than 60 languages. Pricing is listed at $0.12 per hour for the streaming interface and $0.10 per hour for the async interface. The item was cited by Techub News, with MarkTechPost named as the source in the brief.