OpenAI's GPT-Live: A voice feature for ChatGPT that can listen and engage in conversation.
OpenAI encourages you to engage with ChatGPT through conversation rather than text. On July 8, it introduced GPT-Live, an advanced generation of voice models. The company claims these models make interactions with AI feel much more like genuine conversations. Starting today, ChatGPT users globally can access two versions: GPT-Live-1 and the smaller GPT-Live-1 mini.
The main update lies in the architecture. GPT-Live operates in “full-duplex,” allowing it to listen and speak simultaneously. It can insert responses like “mhmm” or “got it,” facilitate quick exchanges, or allow pauses for thought. Previous voice models had to wait for you to finish speaking, which often led to interruptions at inappropriate times.
Maintain your thoughts while conversing
The second change involves delegation. When a query requires a web search or more complex reasoning, GPT-Live refers it to a more powerful model running in the background. At launch, this model is GPT-5.5. While waiting for the response, GPT-Live continues the dialogue, seamlessly integrating the answer when it arrives.
“This mirrors human interaction perfectly,” ChatGPT voice product lead Atty Eleti explained to reporters, as noted by CNET. “We keep the conversation advancing while background processing occurs.”
A noteworthy demonstration occurred during press briefings, where the model provided real-time, simultaneous translation, delivering a running translation as the presenter spoke. TechRadar, which was shown this capability, described it as truly beneficial. The model also responds to a designated wake word, with OpenAI staff using “Hey Chat” during the demonstration.
Enhanced, and accessible to all
OpenAI describes GPT-Live as its most intelligent voice model to date. Users can select their desired level of reasoning: Instant, Medium, or High. The voice now has the capability to show visual cards for items such as weather, stocks, and sports. The company asserts that both GPT-Live-1 and the mini outperformed the previous Advanced Voice Mode in conversational tests lasting five to ten minutes. Additionally, significant improvements were noted in benchmarks relating to scientific reasoning and web search capabilities.
According to OpenAI, over 150 million individuals use ChatGPT’s Voice and Dictation features weekly. GPT-Live-1 will be the default option for Go, Plus, and Pro subscribers, while free users will have access to the mini version. The rollout will be available on iOS, Android, and the web, though it may take a few days to reach all users.
The challenge of sounding human
Achieving a human-like voice presents both advantages and challenges. OpenAI has implemented voice-specific safety training and real-time protections to guide responses, provide crisis resources, or terminate conversations in higher-risk situations, with additional safeguards for teenagers. The company emphasizes that GPT-Live utilizes preset voices and does not emulate real individuals.
This caution is important. A model that resembles a friend can gain trust more easily while also posing a greater risk of misleading users. Critics, such as Meredith Whittaker from Signal, caution that chatbots “are not your friends.” The competitive landscape is also intense, with rivals like ElevenLabs and numerous voice-agent startups pursuing similar goals, prompting Amazon to revamp Alexa.
The significance of this development
This launch coincides with OpenAI's ongoing implementation of a wider GPT-5.6 rollout, part of a rapid release strategy. Historically, voice has been considered a convenient supplement rather than the primary method of using ChatGPT. With GPT-Live, OpenAI is betting that spoken interaction will become the norm. However, the question remains whether users will embrace an AI that is always listening.
Other articles
Your complimentary mobile VPN is a privacy nightmare. Who would have thought?
Researchers evaluated 281 free Android VPN applications and discovered unencrypted data transfers, traffic leaks, and extensive tracking within tools that are meant to safeguard users’ privacy.
Meta may be developing smart glasses equipped with AI super sensing capabilities that would allow for constant visual and auditory perception.
Meta is said to be experimenting with "super sensing" AI glasses that will gather extensive information about your life and provide it to an AI assistant that will support you.
Just 6% of managers in the UK believe that Gen Z is prepared for the workforce.
A CMI report reveals that 45% of UK Gen Z believe they are prepared for work, yet only 6% of managers concur, and among those managers, 82% were elevated to their positions without any training.
The GhostApproval flaw disrupts six leading AI coding agents.
Wiz's 'GhostApproval' employs an old symlink method to enable six AI coding agents, including Amazon Q and Cursor, to write beyond the sandbox and transfer control of the box.
CEO of ElevenLabs discusses $600 million in revenue and the AI laboratories.
During the RAISE Summit in Paris, ElevenLabs CEO Mati Staniszewski spoke about achieving a reported $600 million in revenue and outlined the startup's strategies to surpass the AI labs.
Apple begins testing more affordable Chinese RAM in iPhones, but your wallet won't benefit from the savings.
Apple is evaluating memory chips from CXMT, a Chinese chip manufacturer that is supported by the government and is on the Pentagon's watchlist. Washington has previously opposed this and is expected to do so again.
OpenAI's GPT-Live: A voice feature for ChatGPT that can listen and engage in conversation.
OpenAI has introduced GPT-Live, a full-duplex voice feature for ChatGPT that can listen and speak simultaneously, complete with live translation. This feature is being made available to all users, including those with free accounts.