OpenAI Advances Audio AI and Voice Hardware Plans

OpenAI is accelerating development of advanced audio artificial intelligence models, with plans to release a new architecture in the first quarter of 2026, as the company prepares to launch a series of voice-focused hardware devices.

According to multiple sources familiar with the matter, the upcoming model aims to address shortcomings in OpenAI's current audio systems, which trail its text-based models in accuracy, speed, and naturalness. Researchers at the company have noted that while ChatGPT's voice interface is available, few users engage with it, preferring text interactions. To bridge this gap, OpenAI consolidated teams from engineering, product, and research divisions over the past two months into a unified initiative. This effort is led by figures including Kundan Kumar, a voice AI expert recently hired from Character.AI, along with Ben Newhouse on audio infrastructure and Jackie Shannon managing multimodal features for ChatGPT.

The new audio model promises several enhancements, including more emotionally expressive and natural-sounding speech, deeper and more precise responses, and improved handling of real-time conversations. It is designed to manage interruptions...