OpenAI announced the rollout of two new conversational voice models—GPT‑Live‑1 and GPT‑Live‑1 mini—at a press briefing on Tuesday. Both models operate in full‑duplex mode, meaning they can listen and speak at the same time, a capability the company says will let users interrupt naturally and keep conversations flowing without the stilted pauses that have plagued earlier voice assistants.
GPT‑Live‑1, the larger of the pair, will be available to ChatGPT subscribers on paid tiers. The mini version, a stripped‑down variant, replaces the existing Advanced Voice Mode as the default for all users. OpenAI explained that the new models forward each spoken query to its latest text engine, GPT‑5.5, which handles search, reasoning and agentic tasks before the response is spoken back to the user.
Beyond smoother turn‑taking, the upgrade introduces several new features. The voice mode can now stay silent for extended periods, absorbing context until prompted, which the company demonstrated by having the assistant listen quietly while a user walked for up to 40 minutes. Live translation is also supported; a demo showed Hindi translation, though the output retained a pronounced American accent and a slightly formal tone.
OpenAI highlighted visual output as another differentiator. When the new voice mode taps into GPT‑5.5, it can supplement spoken answers with on‑screen graphics, charts or other visual aids, echoing moves by startups such as Monogram, which raised $40 million to blend visual elements into AI responses.
Safety measures received a mention as well. The models include age‑appropriate filters and can provide resources if a conversation drifts toward self‑harm or other sensitive topics. OpenAI stressed that the voice feature is not intended as a companion but as a hands‑free interface for complex, long‑running tasks.
Industry rivals are racing to close the conversational gap. Apple and Amazon have recently upgraded Siri and Alexa with better context handling, while newer entrants like Sesame—founded by Oculus co‑founder Brendan Iribe—are releasing assistants that blend natural dialogue with background task execution. OpenAI’s announcement positions it firmly in this competitive push toward more expressive, human‑like voice AI.
According to the company, more than 150 million people already interact with ChatGPT via voice or dictation. The new models, designed for "most spoken languages," aim to broaden that reach, though OpenAI did not specify the full language list. No hardware, such as AI‑enabled earbuds, was disclosed, despite speculation that the firm might explore such products later this year.
Este artículo fue escrito con la asistencia de IA.
News Factory APP - noticias agénticas para impulsar tu SEO y AEO.