OpenAI is reportedly developing a new voice AI model for voice-based AI devices



OpenAI, a major AI company, is betting big on voice AI, integrating its engineering, product, and research teams to revamp its voice AI, The Information reported. OpenAI's voice AI efforts are said to be in preparation for the company's upcoming voice device.

OpenAI Ramps Up Audio AI Efforts Ahead of Device — The Information
https://www.theinformation.com/articles/openai-ramps-audio-ai-efforts-ahead-device

OpenAI bets big on audio as Silicon Valley declares war on screens | TechCrunch
https://techcrunch.com/2026/01/01/openai-bets-big-on-audio-as-silicon-valley-declares-war-on-screens/

Report: OpenAI plans to launch new audio model in the first quarter - SiliconANGLE
https://siliconangle.com/2026/01/01/report-openai-plans-launch-new-audio-model-first-quarter/

OpenAI merges audio teams, targets new voice architecture by March 2026
https://www.implicator.ai/openai-merges-audio-teams-targets-new-voice-architecture-by-march-2026/

OpenAI's AI Pen Could Turn Your Scribbles Into ChatGPT Gold - Phandroid
https://phandroid.com/2025/12/31/openais-ai-pen-could-turn-your-scribbles-into-chatgpt-gold/

With the emergence of smart speakers such as Google Nest and Amazon Echo , voice assistants have become increasingly popular, with market research showing that more than one-third of American households use voice assistants.

Meta has announced a feature for its AI glasses , such as Ray-Ban Meta, that uses a five-microphone array to pick up on conversations in noisy rooms.

Google is announcing 'Audio Overview' in June 2025, which will convert search results into conversational summaries.

Google is developing an AI feature that will summarize search results in a radio-like voice - GIGAZINE



Tesla is developing Grok, an xAI chatbot, into its vehicles as a conversational voice assistant that can handle everything from navigation to climate control.

It turns out that the chat AI 'Grok' will be standard equipment in Tesla cars delivered after July 12, 2025, and existing vehicles will also be supported by updates - GIGAZINE



In the past, the company also released 'Humane AI Pin,' a voice AI device that could project images onto the palm of your hand, but sales and service for this device have been discontinued.

Humane discontinues Ai Pin and sells business to HP, online functions will be shut down at the end of February 2025, and all data will be lost - GIGAZINE


by Ged Carroll

As major AI companies continue to develop voice AI, it is expected that future voice AI devices will not be 'devices with displays,' but rather devices that project images onto the palm of your hand, in your home, in your car, or anywhere else.

OpenAI is also reportedly focusing on developing voice AI for voice AI devices. This new voice model, scheduled for release in early 2026, will have a more natural voice, respond to interruptions like a real conversation partner, and even speak while the user is still speaking, achieving features not possible with existing voice models.

OpenAI's existing flagship speech model, GPT-realtime, uses the Transformer architecture, but it is unclear whether the new speech model will use a completely different architecture or remain Transformer.

Some Transformer-based voice models process audio directly, while others, like Whisper , released by OpenAI in 2022, convert audio files into a graph called a spectrogram and then process it. OpenAI offers voice models such as Whisper in multiple editions with different output qualities, and it has been pointed out that future voice models may also offer multiple versions.

OpenAI is developing new voice models by integrating its engineering, product, and research teams, led by former Character.AI researcher Kundan Kumar, The Information reported. Character.AI signed a non-exclusive agreement with Google in August 2024, which will see many of the company's co-founders move to Google.

Character.AI co-founders move to Google, Google signs non-exclusive agreement to use Character.AI technology - GIGAZINE



OpenAI is reportedly planning to develop glasses-type and screenless smart speakers, and has already created prototypes of the AI devices it is developing, aiming to ship 100 million units in the fastest time in history .

OpenAI's AI hardware prototype is ready and expected to be completed within two years - GIGAZINE



OpenAI's AI device, internally codenamed 'Gumdrop,' was originally manufactured by Chinese electronics manufacturer Luxshare , but technology leaker @zhihuipikachu reports that it is likely to be transferred to Foxconn due to US import restrictions on China.




Related Posts:

in AI, Posted by logu_ii