OpenAI opens its voice model GPT-Live-1 to developers

OpenAI has released GPT-Live-1 in the API, so developers can now build with the same full-duplex voice model that powers ChatGPT Voice.

The model can listen while it talks, handle interruptions and background noise, and adjust tone and pace along the way. It can also hand off the reasoning itself to another model if the task requires heavier thinking.

The price is $0.05 per minute, affordable enough that it's not only the big tech companies that can experiment with talking AI.

For private users, this doesn't change anything right now. You'll still just encounter ChatGPT Voice as usual.

For companies, it's more interesting: if you're considering an AI answering service, a customer service bot or another voice-controlled solution, it has now become significantly easier and cheaper to build it yourselves.

Rule of thumb: no reason to act right away, unless voice AI is already on your wish list. In that case, it's worth looking into more closely now that access has opened up.

What it means for you

As a private user, you don't need to do anything: you'll still just use ChatGPT Voice as normal. If you work in development, or your company is considering a talking AI solution such as customer service or a phone bot, it has now become both cheaper and easier to get started. If voice AI hasn't been on your radar, it may be worth looking into what it can be used for.