🎬 LookalikeCeleb

Find your celebrity twin with AI  •  Open source  •  Powered by Inference.net

---

📸 How it works

1. Upload a photo (PNG/JPG, paste, or camera)
2. AI analysis via Inference.net Vision + LLM returns structured JSON with your top celebrity matches
3. Results display with hi-res images, similarity scores, and shareable cards
4. Privacy-first - no accounts, no storage, everything processed in memory

🔧 Features

• Transparent AI - structured JSON responses from Inference.net, not black-box results
• Fast development - full-stack multimodal demo in minutes with pnpm run dev
• Type-safe - JSON schema enforcement with TypeScript + Zod validation
• Share-ready - built-in card generator for social media (copy, download, platform links)
• Production-ready - Bun + Hono backend, Docker support, no vendor lock-in
• Privacy-focused - optional analytics, no data storage

🧰 Tech Stack

• Frontend – Vite • React 18 • shadcn/ui • TailwindCSS
• Backend – Bun runtime • Hono router • TypeScript end‑to‑end
• AI – Inference.net Vision API + Structured Outputs
• Deploy – Railway, Vercel, Fly.io, or any Docker host

🚀 Quick Start

git clone https://github.com/yourrepo/lookalikeceleb.git
cd lookalikeceleb
pnpm install                 # or bun install / npm i
cp .env.example .env         # add INFERENCE_API_KEY
pnpm run dev                 # frontend at http://localhost:5173
bun run server:index.ts      # backend at http://localhost:3000

Tip: Vite proxy is pre-configured — uploads hit /api on port 3000 automatically.

📊 Analytics (Optional)

LookalikeCeleb includes optional Plausible Analytics for privacy-friendly tracking:

// In App.tsx - automatically skips if env vars not set
const plausible = Plausible({
  domain: import.meta.env.VITE_PLAUSIBLE_DOMAIN,
  apiHost: import.meta.env.VITE_PLAUSIBLE_HOST,
  trackLocalhost: false, // Only tracks in production
});

Setup (Optional)

Add to your .env file:

VITE_PLAUSIBLE_DOMAIN=yourdomain.com
VITE_PLAUSIBLE_HOST=https://plausible.io
# Or use your own Plausible instance

Remove Analytics

Simply delete the useEffect block in src/App.tsx or leave env vars unset.

Note: Analytics only tracks in production (ignores localhost). No tracking = no data collected.

🏗️ Architecture

flowchart TD
  A[Client React] --> B[API Matches Hono+Bun]
  B --> C[Inference.net Vision LLM]
  C --> B
  B --> D[Image Search Proxy]
  B --> A

Loading

Flow:

• Client uploads image to Hono API
• Server sends vision prompt to Inference.net with JSON schema
• AI returns structured celebrity matches
• Server fetches hi-res images via search proxy (avoids CORS)
• Combined response sent back to client

✨ API Example

Inference.net Request → Response

Request

POST https://api.inference.net/v1/chat/completions
{
  "model": "google/gemma-3-27b-instruct/bf-16",
  "messages": [
    {"role":"user","content":[
      { "type":"image_url",
        "image_url":{"url":"data:image/jpeg;base64,..." }
      },
      { "type":"text",
        "text":"Find top 3 celebrity look‑alikes with explanation."}
    ]}
  ],
  "response_format":{
    "type":"json_schema",
    "json_schema":{
      "type":"object",
      "properties":{
        "matches":{"type":"array","minItems":3,"maxItems":3,
          "items":{
            "type":"object",
            "properties":{
              "name":{"type":"string"},
              "percentage":{"type":"number"},
              "description":{"type":"string"}
            },
            "required":["name","percentage","description"]
          }
        }
      }
    }
  }
}

Response

{
  "matches": [
    { "name":"Emma Stone","percentage":94,
      "description":"Wide-set green eyes, pronounced cheekbones…" },
    { "name":"Ryan Gosling","percentage":87,
      "description":"Similar jawline, nose bridge, blue eyes…" },
    { "name":"Zendaya","percentage":82,
      "description":"Matching eyebrow arch, chin profile…" }
  ]
}

🚀 Deploy

Platform	Instructions
Railway
Docker	docker build -t lookalikeceleb . && docker run -p 3000:3000 --env-file .env lookalikeceleb

Set INFERENCE_API_KEY in your environment variables.

🤝 Contributing

We welcome contributions! Ideas for improvements:

• New share card templates
• Additional AI providers
• Performance optimizations
• Dark mode
• Mobile app version

1. Fork this repo
2. Create a feature branch: git checkout -b my-feature
3. Commit your changes and open a PR

📜 License

MIT - feel free to use this in your own projects.

---

A practical demo of multimodal AI for the open source community.
Built to show what's possible with modern vision models.