Lovely, Plan It is a live, camera-first travel companion for the real world.

Instead of forcing travelers to stop, type, compare search results, and manually piece together context, Lovely lets them point the camera, speak naturally, interrupt mid-conversation, and get grounded help in the moment.

1. The problem we solve

Travel discovery is still broken in small but important ways.

Search and maps are good at generic answers, but weak at the moment when someone is actually standing somewhere unfamiliar and asking: Where am I? Is there anything interesting here? How do I get there? Is it worth going now? What did that poster say? When is the next bus?

Existing tools also ignore context: who is traveling, whether they are walking or driving, what they already saw earlier, and what they actually enjoy. As a result, travelers get fragmented, one-size-fits-all answers instead of a live guide that understands place, preference, memory, and next action.

2. How it works as a product

Lovely is a smart travel camera. The user opens the web app on their phone, shares camera, microphone, and location, and talks to Lovely naturally. Lovely can recognize what the traveler is looking at, explain why it matters, answer follow-up questions, and adapt to interruptions in real time. It also remembers what the traveler saw earlier in the day, including posters, menus, visible text, QR codes, and booking links, so those can be revisited later as part of travel memory.

Recommendations are personalized using traveler profiles such as Solo, Family, Date, or Work, combined with the current mobility mode such as walking, driving, or public transport. Lovely can then save places and events into future plans, assign dates, and even email calendar invites. The result is not just a chatbot with vision, but a travel agent that can see, hear, speak, remember, and help the user act.

3. How it works with Gemini and Google Cloud

Lovely is built around the Gemini Live API for real-time multimodal conversation, so the core interaction is direct live voice plus camera context rather than turn-by-turn text chat. Gemini handles live understanding, tool use, grounded reasoning, and fresh re-reading of stored memory images when needed.

On the cloud side, the backend runs on Google Cloud via Firebase App Hosting / Cloud Run, where it mints ephemeral Live tokens, executes tool calls, and orchestrates grounded services. Firestore stores traveler profiles, place history, local intelligence, memories, future plans, and background enrichment jobs.

Firebase Storage stores captured memory media, and Firebase Cloud Messaging is used for push notifications such as area-guide readiness. Cloud Tasks powers autonomous background enrichment jobs that prepare area guides for new destinations.

Lovely also uses Google Maps Platform for Places, Geocoding, and Routes, combining those with Gemini-powered search grounding, local curated place/event intelligence, transit planning, weather, QR extraction, and calendar invite delivery.

This creates a single live system where Gemini is the conversational brain, Google Cloud is the orchestration and data layer, and the phone becomes the traveler’s real-world interface.

Share this project:

Updates