Soniox STT v6 Adds Agent Text Input and Echo Filtering
Soniox released Soniox STT v6 speech-to-text API, introducing agent text input and echo filtering features available in streaming and async versions.
Stock photo for illustration only, not from the actual event
- Soniox released Soniox STT v6, the latest speech-to-text API model.
- Introduces agent text input to recognize short replies effectively.
- Includes echo filtering to drop the voice agent's own speech.
- Available in two versions: stt-rt-v6 for streaming and stt-async-v6.
Soniox has officially released Soniox STT v6, representing the latest generation of its speech-to-text API designed specifically to enhance conversational voice agents and automated phone systems.
The core capability of Soniox STT v6 involves reading what a voice agent says while simultaneously listening to the caller, utilizing that text data to accurately recognize short replies from users.
Stock photo for illustration only, not from the actual event
In addition to text input, the new release features built-in echo filtering designed to drop the voice agent's own voice, preventing feedback loops and confusion during active conversations.
This latest generation is deployed across two distinct models to cater to different operational needs, shipping as stt-rt-v6 for real-time streaming and stt-async-v6 for asynchronous processing.
Integrating direct agent text input into a speech recognition model addresses a major bottleneck in voice AI development. By knowing precisely what the system uttered, the model can filter out self-generated echoes and better parse brief user confirmations without getting confused by audio overlap.
Source: MarkTechPost
Found something wrong in this article? Report an issue with this article
Comments
Leave a Comment