marktechpost.com web signal

Bland AI Unveils Tofu-1, a Full-Duplex Speech-to-Speech Model Built for Noisy Phone Calls

Voice AI Generative AI ai-business

Summary

Bland AI unveiled Tofu-1 on Nov 15 in enterprise research preview — a native full-duplex speech-to-speech model that listens, reasons and speaks simultaneously without a text intermediary. The company claims it can isolate a caller's voice from crying children, TVs and road noise and keeps talking while handling backend tasks, targeting hospital and in-car calls. Tofu-1 is Bland's first in a planned speech-to-speech model family; no benchmarks, size or pricing have been published.