Back to rankings

Lovable: Live Voice Conversations With GPT Live

LovableView original changelog

Lovable now lets builders add live voice conversations to their apps using GPT Live. Users can talk to an app, interrupt it mid-answer, and get real-time spoken responses, which suits conversational tutors and spoken assistants. The feature requires projects built on TanStack Start and consumes credits.

Key Takeaways

  • Live voice conversations can now be added to apps built in Lovable using the GPT Live model.
  • Users can interrupt the app mid-response and receive real-time spoken answers, which makes exchanges feel natural.
  • Voice sessions can trigger actions in the app, not only produce spoken replies.
  • The feature is aimed at use cases such as conversational tutors and spoken assistants.
  • It is available only for projects on the TanStack Start framework.
  • Voice usage consumes credits, so it adds to workspace spend.

Talk to your app in real time

Lovable added support for GPT Live, a live voice model, as a building block for the apps created on its platform. Builders can now give an app a voice interface where people speak, listen, and carry on a natural back-and-forth conversation instead of typing into a text box.

What the conversations can do

Voice sessions are interactive rather than turn-based. A user can interrupt the app while it is speaking, and the app responds in real time. Lovable also notes that conversations can trigger actions in the app, so a voice assistant can do more than answer questions. The examples Lovable points to are conversational tutors and spoken assistants, where a hands-free, natural exchange matters more than a typed prompt.

Requirements and cost

The capability requires projects built on the TanStack Start framework. Voice use consumes credits, so builders adding it to an app should factor the usage into their workspace spend. Lovable documents the feature in its AI features guide under live voice models.

Why it matters

GPT Live is the model family OpenAI introduced for full-duplex voice, where the model can listen and speak at the same time. Bringing the same style of interaction to apps built in Lovable moves voice from a niche add-on to something a builder can describe in a prompt and ship, which widens the kinds of products non-specialist builders can create.