The Hook
“All right, all right guys. So, introducing Yap.”
Product Reveal
“(No spoken dialogue — upbeat lo-fi music)”
Problem Agitation
“Yap is a tool for your agent that gives your agent harnesses the eyes, years, and hands to record, understand, and edit videos locally. It's actually got no AI capability of its own, it's just a set of nice primitives to record and edit videos that you could give to your agent.”
Feature Teaser
“Basically what you do is it's just a very simple prompt that you copy and that will teach your agent harness how to install Yap on your computer. Once it's installed, and it's going to give you a CLI and a set of skills that your agent can use to actually edit and export those videos for you.”
Product Reveal
“Now, whenever I want to do these videos, I just yap into the agent. I just yap just randomly. I don't need to be overly prepared with all these different takes and stuff. I could just, you know, do as many takes as I want and just keep yapping. And when the agent has Yap, it could basically just record stuff from its webcam, or it could import videos from any other camera like what I'm doing here. And then it's also got its set of tools to transcribe and splice videos together.”
Call to Action
“My agent is going to take all these videos, and it will turn it into a nice launch video, just like this one. Just go to justyap.io to install it. There's a prompt you could just give to your agent. The agent will know what to do from there. Go try it out. Thank you so much.”
Feature Teaser
“So Claude, when you're editing this later in the future, you have all these different feeds, you have my screen recording, you have that far camera over there, you have this MacBook camera over here. So Claude, make sure to use all these feeds, and you can cut between these different feeds and the screen recording when you edit this video in the future, okay? Remember, when the AI is editing it, you don't need to do takes, right? It's just one long recording, the AI can crop it however it wants. Stop stop recording, I want to have the agent edit it now.”
Synthesizable in Remotion · key components of the agent-multicam-launch template.
Dynamic 2x2 grid displaying multi-angle synchronized synchronized video sources with labelled pills.
Use CSS Grid with 4 absolute positioned OffthreadVideo components and animated layout transitions via spring().
Playful sticky note cards popping into frame with slight rotational overshoot and handwriting typography.
Wrap children in an animated div driven by spring({ damping: 12, stiffness: 100 }) applying rotateZ and translateY.Monospaced timestamped transcript with highlighting and strikethrough cut indicators.
Map transcript array with frame-interpolated opacity and conditional strikethrough red text.
Purpose. Demonstrate the premise instantly using raw multi-camera feeds.
Execution. Split 4-camera view showing raw synchronization.
Purpose. Clarify that Yap provides primitives without internal model lock-in.
Execution. Warm paper illustrations showing agent toolboxes.
Purpose. Demonstrate self-referential proof that the video itself was edited by an AI agent.
Execution. Raw outtakes giving instructions directly to Claude.
Objective
Launch an AI coding agent CLI that edits source code from voice commands.How the recipe adapts
Replace multi-cam feeds with simultaneous IDE windows, terminal outputs, and webcam reactions, using paper sticky notes to highlight auto-refactored code blocks.Scene durations (4 scenes, 60s)
Key Remotion components
Introducing Griffin, the first AI to pass the video Turing test
The most emotive and controllable TTS model

Always-on agents built to handle everything