Lovable builds GPT-Live, Claude and Gemini Omni into its chat: 9 AI tools in one place

Lovable has expanded the set of AI capabilities that can be embedded in the apps it builds to nine tools from seven providers. They include OpenAI's voice-based GPT-Live, current Claude models from Anthropic, Gemini 4 Argon, Gemini Omni and Lyria from Google, plus Perplexity, Firecrawl, Parallel and Jev from Typesafe AI. The announcement was published on the company's blog on 15 October 2026. The core idea is that the app's author does not pick an API, set up keys or write glue code. They describe the task in chat, and Lovable chooses the model itself.

Voice, reasoning and multimedia

The most notable addition is GPT-Live, a voice mode for real-time conversation. According to Lovable, it can power an assistant that answers questions, guides users through an interface and sees a task through to the end, for example booking an appointment or helping with support. The company cites a language tutor for practicing speech before an exam and a sales coach that asks follow-up questions about a pitch. One employee built an app that lets you talk to Rocky from Project Hail Mary.

Lovable positions the Claude models for tasks that involve a long chain of actions: deep reasoning, planning and a clean final result. Examples include turning a client brief into a polished commercial proposal, appraising an item from a marketplace listing photo, and an email assistant that flags urgent messages and drafts replies.

Google is represented by three products at once. Gemini Omni animates photos and creates video from various inputs. Lyria writes original tracks with lyrics, and the user sets the style, tempo and vocals. Gemini 4 Argon handles analytical work such as reading competitor reports and trend charts. Together they produce apps like a store tool that makes a promo video with music from a single product shot, or a real-estate tool that assembles a video walkthrough from photos of a property.

The web layer and fast triage

The second half of the announcement covers data and routing:

• Perplexity gives answers based on the live web, with citations the user can verify. It suits morning briefings for investors or a city guide.
• Firecrawl searches, scrapes and returns clean data. It fits a competitor price tracker or a help-center assistant that answers from your own documentation.
• Parallel is a web search platform with synthesis. It covers everything from yesterday's news to research such as a map of a company's investments over decades. Use cases include monitoring scientific papers, tracking funding rounds, finding grants and tenders, and lead enrichment.
• Jev from Typesafe AI makes quick decisions: a category, a score, a yes or no. It labels chat messages as a refund, a delivery question or something urgent, and routes tickets to the right team.

The Jev and Claude pairing is telling: the first sorts, the second writes the reply. This is a typical support architecture in which cheap, fast classification sits in front of expensive generation. Lovable packages it into a single prompt sentence. The announcement does not say which models are behind Jev or how pricing works.

How to try it

Lovable suggests not thinking about an "AI strategy" and simply describing a feature, for example: "Add a research panel with competitor summaries and citations." The platform will pick a model or tool automatically. If you want a specific one, you can name it in the chat, such as "use Opus 5.5" or "use Perplexity."

In practice, it makes sense to start with narrow scenarios where the value is obvious right away. For support, that means ticket triage through Jev with draft replies from Claude. For content, it means a video from a product photo via Gemini Omni and Lyria. For analytics, it means a dashboard with cited sources on Perplexity or Parallel. Before launch, check three things: each provider's terms of use, the cost per user, and how the app behaves if an external service is unavailable. The announcement contains no public data on limits or prices, so you will have to confirm them in Lovable's documentation.

Over the past two days, the company has shipped two more updates. On 14 October it added help with finding customers for finished apps, and on 15 October it announced deployment to Azure in the user's own subscription. Together with the new set of models, Lovable is building a full cycle from idea to first users inside a single platform.