WildVerse

Don’t catch creatures. Notice them. Turn real-world wildlife photography into a collectible fantasy companion bestiary powered by Google Gemini.

Comment

Inspiration

Every day, millions of people walk past incredible local biodiversity without noticing it. Meanwhile, millions play games centered around catching and collecting virtual monsters.

We asked a simple question: What if the addictive reward loop of a creature-collecting game could make people pay attention to real-world nature?

WildVerse flips the typical gamification script:

  • Don't catch creatures in nature. Notice them.
  • The fantasy companion is the reward.
  • The biodiversity observation is the real value.

What it does

WildVerse turns real-world wildlife photography into an interactive citizen-science bestiary powered by Google Gemini.

The 5-Step Core Loop

  1. Photograph Wildlife: The user captures or uploads a photo of an animal in their local environment.
  2. Gemini Multimodal Analysis: Gemini 3.6 Flash analyzes the photo and outputs a conservative, structured identification including:
    • Common name & Scientific Latin name
    • Taxonomic group (Mammal, Bird, Amphibian, Insect, Reptile, etc.)
    • Confidence level & observation quality
    • Short educational facts & habitat context
    • Responsible field observation safety notes (e.g. Observe from a distance, never handle or disturb wildlife)
  3. Fantasy Companion Creation: Gemini generates an original collectible fantasy companion inspired by the real animal's silhouette, colors, and ecological niche — strictly avoiding copyrighted character designs.
  4. Native Image Transformation: Gemini 3.1 Flash Image transforms the reference photograph into stylized digital concept artwork for the bestiary companion.
  5. Gamified Progress & Atlas Contribution: The player earns XP, levels up their Field Level, unlocks achievements/badges, and adds the sighting to a living community biodiversity atlas.

How we built it

WildVerse is built on a high-performance FastAPI backend with SQLite persistence and a vanilla HTML5 / CSS3 / JS frontend designed with modern glassmorphism aesthetic standards.

Gemini API Integration (Core Architecture)

Gemini is central to every step of the pipeline:

  • Multimodal Vision (gemini-3.6-flash): Reads raw image bytes to identify species, rate confidence, and synthesize ecological context.
  • Structured JSON Schema Outputs: Enforces strict Pydantic/JSON schemas (response_schema), ensuring reliable, typed API responses that map cleanly into application state.
  • Companion Concept Generation: Synthesizes original names, elements (Forest, Water, Spark, Air, etc.), lore, and abilities grounded in real biological traits.
  • Image Generation (gemini-3.1-flash-image): Transforms animal reference photos into full-body concept art.

Privacy & Open Tech

  • Leaflet & OpenStreetMap: Zero-cost interactive atlas mapping with no paid Google Maps dependencies.
  • Deterministic Demo Mode: Allows judges to experience 100% of the product flows instantly without requiring API keys or setup steps.

Challenges we ran into

  1. AI Certainty vs Scientific Integrity: AI vision models can misidentify species, especially with partial or blurry photos. We solved this by enforcing conservative prompt rules, displaying confidence percentages, using guarded language (likely, possible), and providing clear safety warnings.
  2. Avoiding Copyright Overlap: Ensuring generated fantasy companions felt magical without mimicking trademarked franchise characters (e.g. Pokémon, Digimon). We crafted strict system prompts forbidding Pokémon names, terminology, Pokéballs, or character imitations.
  3. Real-time Perceived Speed: Multimodal vision + structured output + image generation can take several seconds. We implemented animated 3-step progress indicators (Understanding species → Designing companion → Mapping sighting) so the user experience feels swift and continuous.

Accomplishments that we're proud of

  • ✦ Meaningful Gemini Integration: Gemini is not a chatbot addon — it is the sole engine driving both species recognition and creative game design.
  • ✦ Structured Output Reliability: Zero unparsed text. 100% structured JSON outputs powering database records, UI cards, and map markers.
  • ✦ Zero-Cost Production Ready: Built entirely with FastAPI, SQLite, Leaflet/OSM, and Gemini free tier capabilities.
  • ✦ Responsible Citizen Science: Promotes nature appreciation without incentivizing animal displacement, trapping, or harassment.

What we learned

Multimodal AI is most powerful when it bridges perception and motivation. Gemini doesn't just tell users what an animal is — it translates that real-world encounter into an original creative artifact that gives the user a sense of discovery, ownership, and pride in their local ecosystem.

What's next for WildVerse

  • 🌿 Community & Expert Verification: Allow local naturalists to review and verify sightings.
  • 🛡️ Geoprivacy Protection: Automatic coordinate fuzzing for endangered or sensitive species.
  • 🏫 School & BioBlitz Challenges: Group field events for local schools and park networks.
  • 📱 Mobile Native App & Offline Logging: Queue field photos while offline for batch Gemini analysis upon returning to connectivity.

Built With

Share this project:

Updates

Submission history