This is a submission for the Google AI Studio Multimodal Challenge
What I Built
I created The Stargazing App — a multimodal web experience that transforms your phone or laptop into a pocket observatory. By simply entering your location and time, you’ll instantly see what the night sky has in store: from twinkling constellations to planets on the horizon. The app blends astronomy data with AI-powered storytelling, so users don’t just see the stars — they understand them. 🌠
Demo
Check out the live demo here → YouTube Demo
How I Used Google AI Studio
Google AI Studio was the heart of Starlight Storyteller, empowering me to harness Gemini 1.5 Flash’s multimodal prowess for a stellar user experience. Here’s how I wielded its magic:
Model Selection 🔍: I chose Gemini 1.5 Flash in Google AI Studio for its text and image processing capabilities, perfect for blending astronomical data with sky map annotations—all within the free tier, keeping costs at zero! 💸
Prompt Engineering ✍️: In AI Studio’s intuitive interface, I crafted a structured prompt to make Gemini an “AstroGuide.” It analyzes JSON data (e.g., {"planets": [{"name": "Jupiter", "ra": 5.91, "dec": 23.13, "alt": 45.0}]}) and sky map images, delivering summaries like: “Jupiter glows at 45° altitude in New York, named for the Roman god-king!” I iterated in Studio to perfect the tone—fun, educational, under 400 words. 🪐📖
Multimodal Testing 🖼️: I uploaded sample PNG sky maps (generated via Matplotlib) to AI Studio, testing how Gemini annotates objects (e.g., “Jupiter is the bright dot at center”). This ensured accurate image-text integration before coding API calls. 🌟
Export to API ⚙️: Once the prompt shone brightly, I exported it as a tuned model ID for Vertex AI. AI Studio’s Python snippets made integration into my FastAPI backend a breeze, using Image.from_bytes to send sky map PNGs to Gemini. 🚀
Validation & Iteration 🔄: AI Studio’s fast feedback loop let me tweak prompts, test outputs, and refine instructions (e.g., handling unclear images by relying on JSON). This slashed debugging time when deploying to Cloud Run. 🛠️
Multimodal Features 🌍📷
Text + Image Fusion: Users provide location/time, and the app generates both a written summary and annotated visuals.
AI-Powered Storytelling: Gemini narrates myths, facts, and tips, turning cold star charts into cosmic stories.
Dynamic Visuals: A PNG sky map is annotated by Gemini, while a D3.js-powered SVG lets users interactively explore the sky.
Smart Reliability: Caching ensures repeat lookups are lightning-fast ⚡, and offline fallbacks mean you’re never lost under the stars.
(https://dev-to-uploads.s3.amazonaws.com/uploads/articles/eij3khdystsuenhit9lv.png)
Final Thoughts 🌌💡
This project taught me how multimodal AI can transform raw data into meaningful human experiences. With Google AI Studio as my creative lab and Gemini as my storytelling partner, I built not just an app — but a stargazing companion for dreamers, learners, and explorers everywhere. 🌙✨
Top comments (0)