OpenAI finally rolls out Advanced Voice Mode to paid ChatGPT users

After months of demos, delays, and voice actor drama, OpenAI is finally opening the floodgates.

Sam Altman announced that Advanced Voice Mode (AVM) began rolling out to all ChatGPT Plus and Team subscribers, with access completing over the week.

OpenAI even poked fun at the long wait, noting the updated model can say "Sorry I'm late" in over 50 languages.

What made the cut (and what didn't)

When OpenAI first teased the hyper-realistic voice features in May, it blew minds with instant responses, natural laughs, and emotional inflections. A limited alpha went out in July, but most paying users were left waiting.

During that downtime, OpenAI baked in two crucial features: Custom Instructions and Memory. ChatGPT can now remember your personal details and conversational preferences across spoken chats.

They also added five new nature-inspired voices while improving accent recognition and response latency. Unsurprisingly, the controversial "Sky" voice that sounded strikingly like Scarlett Johansson remains scrapped.

Regional restrictions are the biggest catch. Subscribers in the EU, the UK, Switzerland, Iceland, Norway, and Liechtenstein are completely excluded from this rollout due to regulatory hurdles.

More importantly, the magic demo features are missing. There is no live video or screen sharing, no image upload option during voice chats, and no integration with the new o1 reasoning models yet.

If you are a Plus or Team subscriber outside the blocked regions and still do not see the feature, try uninstalling and reinstalling your ChatGPT mobile app.

Why it matters

We think this rollout marks a major shift in how people actually interact with artificial intelligence on a daily basis.

For everyday tasks like language practice, brainstorming aloud during a commute, or quick storytelling, AVM makes talking to a computer feel conversational rather than transactional. The latency is practically gone, and being able to interrupt mid-sentence without triggering an error changes the entire vibe.

At the same time, launching without vision or o1 reasoning feels like a calculated compromise. OpenAI needed to give paying subscribers a win after months of anticipation, even if it meant shelving the multimodal "Her" experience for later.

It sets a high bar for Google Gemini and Apple Intelligence. If native smartphone assistants cannot match this level of fluid, interruption-friendly voice interaction soon, OpenAI could easily become the default voice interface on your phone.