In the last two months, OpenAI has unified several engineering, product and research teams with a single objective: to revolutionize its audio models.
The startup is preparing a more natural voice model for this first quarter of 2026, capable of managing interruptions and speaking while you speak, according to a report from The Information.
Why is it important. This movement not only seeks to improve ChatGPT, but also to place audio as the main interaction interface, moving screens to the background at least in certain use cases. This is what first-generation smart speakers tried, unsuccessfully, a decade ago.
The bet is to build personal devices that work exclusively by voice, with a launch planned for mid-2027.
The context. Silicon Valley has been heading in this direction for months:
- Meta added five microphones to his Ray-Ban Meta 2 to isolate voices in noisy environments.
- Google is testing audio search summaries.
- Tesla is going to integrate Grok in their cars to be able to control certain aspects conversationally.
In detail. The initiative is led by Kundan Kumarformer researcher of Character.AI which arrived at OpenAI this summer. The new model seeks to sound indistinguishable from a human voice and maintain fluid conversations without the typical cuts of current assistants.
Besides, the May 2025 purchase of io Products Inc.Jony Ive’s $6.5 billion startup, marks a turning point. Ive, former head of design at Apple, now leads creative responsibilities at OpenAI with a team of 55 people. Its philosophy, already publicly announced, seeks to reduce addiction to devices through interfaces that do not require constant visual attention.
What is happening. OpenAI contemplates several formats: screenless speakers, smart glasses (a clearly booming segment) and a pen-shaped, voice-operated device.
Foxconn will manufacture the first product, rumored to be a context-aware pen, in Vietnam. These devices are positioned as complements to laptops and mobile phones, not as substitutes, at least for now.
Yes, but. Not all “screenless AI” bets have worked.
- The Humane AI Pin burned hundreds of millions and defrauded its buyers by offering a half-hearted product that would stop working after the company was sold to HP.
- Several pendants have been in a similar line for almost two years, without any to date having managed to go beyond being a curiosity.
And now what. The schedule is quite tight:
- New audio model before spring 2026.
- First dedicated device for sale a year later.
OpenAI will go from being a software provider to competing directly in consumer electronics. The question is whether they will achieve what Humane and others have failed to achieve: make people want to talk to their devices without being able to look at a screen.
In Xataka | The new Ray-Bans from Meta will allow you to cross a line: seem present while you are completely absent
Featured image | Xataka with Mockuuups Studio
GIPHY App Key not set. Please check settings