Inspiration

Extraordinary moments happen every minute. A lightning strike over the ocean, a meteor crossing a city skyline, a once-in-a-lifetime wildlife encounter. Yet the people who capture them on their phones receive nothing. Meanwhile, social media platforms profit from viral content while creators generally earn zero. I wanted to build a system that flips this dynamic: one where the rarity of a moment is objectively scored by AI and instantly rewarded on-chain, no middlemen, no negotiation, no waiting for monetization thresholds. Clipchain was born from a simple belief: if you witnessed something extraordinary, you deserve to be paid for it.

What it does

Clipchain is a decentralized video bounty network where users upload rare video clips and earn SOL automatically based on AI-assessed rarity. It has four core features:

Rarity scoring engine — TwelveLabs indexes and semantically understands each clip's content, while Gemini API classifies the scene and estimates global frequency. Combined, they produce a 0–10 rarity score using the formula SOL = rarity² × 0.025 × demand_coeff Instant SOL payouts — clips scoring above 6.0 trigger a Solana smart contract that releases a bounty directly to the uploader's wallet, with no manual review required

NFT minting — clips scoring above 8.5 are automatically minted as NFTs via the Metaplex program, with TwelveLabs semantic embeddings stored as metadata, and listed on an auction marketplace for collectors Live processing dashboard — users can track their clip through the full pipeline in real time: Vultr upload, TwelveLabs indexing, Gemini classification, rarity calculation, and payout confirmation, all within seconds

How we built it

Our stack combines AI, blockchain, and cloud infrastructure into a single seamless pipeline: Frontend: Built with Next.js and TypeScript, styled with Tailwind CSS. Wallet connection is handled via the Phantom and Solflare SDKs — no email or password required, just a Solana wallet. Video Storage: Vultr Object Storage receives clips directly from the browser via presigned URLs, so footage never routes through our backend servers. A Vultr GPU compute instance (NVIDIA A100) runs the orchestration layer that coordinates AI calls in parallel. AI Processing: TwelveLabs handles deep video understanding — identifying objects, actions, motion patterns, and scene context to produce a semantic embedding. Gemini API independently classifies the scene category and estimates rarity based on world knowledge. Both calls run concurrently on the Vultr compute layer and feed the scoring formula. Blockchain: Solana smart contracts hold a bounty escrow pool. When the rarity engine pushes a verified score and oracle signature, the contract computes the payout and releases SOL atomically. NFT minting is handled through the Metaplex standard with TwelveLabs embeddings stored as on-chain metadata. Backend API: A lightweight Node.js server on Vultr brokers between the frontend, AI providers, and the Solana program. Push notifications alert wallets when processing completes.

Challenges we ran into

Parallel AI orchestration: Coordinating TwelveLabs and Gemini calls concurrently without race conditions in the scoring pipeline required careful async design and result validation before triggering the smart contract Rarity calibration: Defining what "rare" actually means numerically was harder than expected — early versions scored sunsets as highly as lightning strikes. I iterated on the scoring formula and Gemini prompt engineering to create meaningful differentiation across event categories Presigned URL uploads at scale: Handling large video files (up to 500MB) directly from the browser to Vultr without timeouts required chunked multipart uploads and resumable upload logic Oracle security: Ensuring the smart contract only accepted legitimate rarity scores — not spoofed requests — required building a signed oracle layer with wallet-bound verification

Accomplishments that we're proud of

Built a fully end-to-end pipeline — from clip upload to SOL in wallet — in under 36 hours Achieved sub-15-second total processing time from upload completion to payout confirmation Designed a wallet-native UX with zero traditional sign-up friction Successfully deployed and triggered a real Solana smart contract payout on devnet during the demo Created a dual AI analysis system where TwelveLabs and Gemini cross-validate each other, reducing false high scores Built an NFT metadata standard that stores semantic video embeddings — not just a file pointer — making each token genuinely descriptive of the captured event

What we learned

How TwelveLabs' video embedding model works under the hood and how to query it for scene-level classification versus action-level understanding. How to architect presigned URL direct-upload flows that bypass backend bottlenecks for large media files The nuance of designing AI scoring systems that are resistant to gaming — and how quadratic reward curves create better incentive structures than linear ones That "wallet as identity" is not just a Web3 philosophy, it genuinely simplifies the auth layer for developer and user alike

What's next for Clipchain

  1. Rarity model improvements

Per-category rarity baselines (a meteorological rarity scale vs. a wildlife one) for fairer scoring Temporal rarity — a lightning strike captured twice in one day is less rare the second time Community dispute resolution for borderline scores

  1. Marketplace and collector economy

Collector wishlists that boost the demand_coeff for specific event categories, dynamically raising bounties for sought-after clips Royalty splits on secondary NFT sales flowing back to original uploaders Curated "event collections" grouping related rare clips into bundle NFTs

  1. Mobile-first capture

A native iOS and Android app with direct in-camera upload, so users can submit clips seconds after capturing them Snapdragon on-device pre-screening to filter obviously low-rarity clips before wasting upload bandwidth

  1. Decentralized governance

A DAO where SOL stakers vote on rarity thresholds, bounty formula parameters, and new event categories Transition the oracle layer to a decentralized validator network so no single backend controls payout eligibility

Our vision is for Clipchain to become the world's largest archive of verified rare moments — a living record of the extraordinary, owned by the people who witnessed it.

Built With

  • react
  • solana
  • twelvelabs
  • vultr
Share this project:

Updates