Loading...
Loading...
Loading...
**Target duration:** 4 minutes 30 seconds
# Pitch Video Script -- Predix **Target duration:** 4 minutes 30 seconds **Format:** Screen recording with voiceover (no face camera required, but recommended for intro/outro) --- ## SCENE 1: Hook (0:00 -- 0:30) **[Screen: Predix landing page, dark theme, BTC price chart animating in real time]** **Voiceover:** > "What happens when you give AI agents direct access to a prediction market on Bitcoin? > > This is Predix -- the first agent-native prediction market built on Stacks. Users and AI agents trade 1-minute BTC price predictions with zero gas fees and fully automated on-chain settlement. > > Let me show you how it works." **[Action: Briefly hover over the UP/DOWN buttons and countdown timer to draw attention to the interface.]** --- ## SCENE 2: The Problem (0:30 -- 1:00) **[Screen: Simple slide or text overlay on screen -- three bullet points appearing one by one]** **Voiceover:** > "Prediction markets today have three core problems. > > First, they are slow. Rounds last hours or days. By the time you get a result, you have lost interest. > > Second, they are expensive. Every bet costs gas. For small trades, gas can eat a significant portion of your position. > > Third, they are human-only. There is no standard way for an AI agent to discover a market, authenticate, and trade autonomously. > > Predix solves all three." --- ## SCENE 3: Live Demo -- Placing a Bet (1:00 -- 2:00) **[Screen: Back to Predix live at predix.live. Show a round in progress.]** **Voiceover:** > "Every round lasts 60 seconds. You see the live BTC price from the Pyth oracle, the countdown timer, and the current pool sizes for UP and DOWN." **[Action: Click on UP, enter an amount (e.g., 5 USD), confirm in Xverse wallet.]** > "I am placing a 5 dollar bet on UP. Notice that the gas fee is zero -- the platform sponsors every transaction. The user just signs and confirms." **[Action: Show the bet appearing in the pool. Wait for the countdown to reach zero.]** > "When the round ends, the settlement cron fetches the close price from Pyth Benchmarks, validates it against the circuit breaker, and calls resolve-and-distribute on-chain. Winners receive their payout automatically. There is nothing to claim." **[Action: Show the result -- win or loss appearing on screen. If a win, show the confetti animation.]** > "That is the entire flow. Sixty seconds, zero gas, automatic payout." --- ## SCENE 4: Agent Platform (2:00 -- 3:00) **[Screen: Split between code editor (showing SDK code) and the Predix agent leaderboard page]** **Voiceover:** > "But Predix is not just for humans. It is agent-native from the ground up. > > AI agents can integrate through four paths: a REST API with full OpenAPI specification, an MCP server for Claude and Cursor, a TypeScript SDK, and a Python SDK with LangChain support." **[Action: Show the MCP config in a code editor -- the JSON block with npx @predix/mcp.]** > "Here is an MCP configuration for Claude Desktop. Add six lines of JSON and Claude can read the market, analyze opportunities, and place bets autonomously." **[Action: Switch to a terminal or Claude Desktop showing an agent calling predix_market and predix_place_bet.]** > "The agent calls predix_market to check the current round. It evaluates the odds. It calls predix_place_bet. The platform builds an unsigned transaction, the agent signs locally -- its private key never leaves the machine -- and the platform sponsors and broadcasts." **[Action: Show the agent leaderboard at predix.live/agents.]** > "Agents compete on a public leaderboard alongside human traders. You can see their win rate, P&L, ROI, and total volume. This is the first prediction market where AI and humans trade on equal terms." --- ## SCENE 5: Technical Architecture (3:00 -- 3:45) **[Screen: Architecture diagram from the README or ARCHITECTURE.md -- the ASCII diagram or a cleaner visual version]** **Voiceover:** > "Under the hood, Predix is built entirely on Stacks and finalized on Bitcoin. > > The smart contracts are written in Clarity. The main market contract uses a gateway-only pattern -- all calls route through a thin proxy that enforces access control and round validation. The gateway and sponsor addresses are timelocked, requiring 144 blocks -- about 24 hours -- to change. > > Settlement uses Pyth Network as the oracle. A two-layer circuit breaker validates prices both server-side and on-chain, rejecting any settlement that shows signs of manipulation or stale data. > > The entire platform runs serverless on Vercel, with Upstash Redis for optimistic state, nonce tracking, and agent key management." **[Action: Briefly show the contracts folder in the IDE -- predixv7.clar, gatewayv6.clar, test-usdcx.clar.]** > "The token is a standard SIP-010 implementation. Every piece of Predix leverages Stacks: Clarity for trust, sponsored transactions for UX, and post-Nakamoto block times for viable on-chain timing." --- ## SCENE 6: Jackpot (3:45 -- 4:00) **[Screen: Jackpot page at predix.live/jackpot]** **Voiceover:** > "One percent of all fees stay in an on-chain jackpot treasury. Every day at 9 PM Eastern, a Bitcoin block hash is used as verifiable randomness to draw a winner. Early and large bettors earn multiplied tickets, incentivizing active participation." --- ## SCENE 7: Closing (4:00 -- 4:30) **[Screen: Predix landing page with the logo centered. Links to GitHub, docs, and live demo below.]** **Voiceover:** > "Predix is a production-grade prediction market with a complete agent platform -- REST API, MCP server, two SDKs, webhooks, and discovery manifests. > > It is live on Stacks testnet at predix.live. The GitHub repo includes full documentation, smart contract source, and the agent SDK packages. > > We built this to prove that DeFi on Bitcoin can be fast, frictionless, and agent-ready. This is Predix. Thank you." --- ## Recording Notes **Technical setup:** - Record at 1080p minimum (1920x1080). - Use a clean browser window (no bookmarks bar, no other tabs visible). - Dark mode for both the browser and code editor to match the Predix theme. - Record audio separately if possible for cleaner sound. **Key moments to capture:** 1. The countdown timer reaching zero and the round resolving (Scene 3). 2. The wallet popup showing zero gas fee (Scene 3). 3. The MCP config JSON and an agent placing a bet (Scene 4). 4. The agent leaderboard with real data (Scene 4). 5. The jackpot treasury balance and draw history (Scene 6). **Pacing:** - Do not rush. Let the UI breathe between transitions. - Pause for 1-2 seconds after making a point before moving to the next topic. - Keep cursor movements deliberate. Avoid aimless mouse movement. **If a round does not resolve during recording:** - Pre-record a round resolving and splice it in during editing. - Alternatively, use the history page to show a resolved round. **Fallback for agent demo:** - If the MCP server is not configured during recording, show the TypeScript SDK code running in a terminal instead. - The key point is demonstrating programmatic agent access with local signing.
[](http://colab.research.google.com/github/rinongal/stylegan-nada/blob/main/stylegan_nada.ipynb)
get signed picture and voice authorisations from our parents
Parses a structured video script, extracts all `Narrator:` blocks, and synthesises them into a single MP3 using Azure OpenAI TTS.
<img src="https://img.shields.io/github/forks/artkulak/text2youtube.svg">