
Bland AI unveils speech model capable of generating multi speaker soundscapes
Bland AI has announced Tofu 1, a voice model engineered to generate complex acoustic environments with multiple overlapping speakers and background ambience. As a full duplex system, it can process incoming audio while simultaneously speaking and filtering out external background noise. Internal tests indicated that most listeners could not distinguish the generated audio from real human voices.
The Blend
Automated phone customer service may soon sound far more human and responsive. Voice technology startup Bland AI introduced Tofu 1, an artificial intelligence system designed to handle phone calls in noisy, unpredictable environments. Unlike traditional customer service bots that process audio by first turning it into text, this system listens directly to raw sound, allowing it to handle mid sentence interruptions, background noise, and multiple people speaking at once.
For everyday callers, this technology could reduce the frustration of dealing with rigid automated phone menus. Standard phone bots often break down if a television plays in the background or if a caller pauses to speak to a child in the room. Bland AI claims its new model can figure out which sounds are directed at the agent and ignore background distractions, meaning calls to banks or emergency services could proceed smoothly even during chaotic real world situations.
However, the technology remains in a limited preview phase for select enterprise clients, so its performance in widespread deployment is still unproven. Furthermore, as voice models gain the ability to master realistic background environments, standard security procedures for phone verification will likely need a complete overhaul to prevent automated impersonation.
Written independently by AI News Smoothie from the reporting listed below. Facts belong to the original publishers. Follow the links for their full coverage.
Ingredients
- Tofu-1: speech-to-speech, built for high-stakes calls | Bland
Bland AI launched Tofu 1, a voice model designed to process live call audio directly and handle overlapping speech during noisy phone calls.