Speech recognition that filters out a voice agent's own speech
The Hook
“For years, we've taught machines to hear. But they could only hear us. What if they could hear themselves?”
Problem Agitation
“Hi! Introduce yourself. Hi there! I'm Nova, your personal— Oh, hi Nova. Nice to— What the... Oh, sorry, go ahead. Stop. No no, you go ahead. Stop. I don't know what you mean by top. Oh, come on.”
Problem Agitation
“Speech recognition can't keep track of what the agent is saying, and it's been breaking the experience for years.”
Product Reveal
“Until now. Introducing STT v6.”
Feature Teaser
“Now with agent text input and built-in echo cancellation.”
Feature Teaser
“Oh, and all of these other improvements.”
Call to Action
“(No spoken dialogue — ambient ocean wave roar and synthetic swell)”
Synthesizable in Remotion · key components of the voice-ai-engine-launch template.
Pulsing gradient sphere responding dynamically to agent speech input and latency lag.
Render an animated radial-gradient div in Remotion modulated via spring physics driven by simulated frame audio amplitude.
A 3D cylindrical carousel showcasing video snapshots revolving smoothly around central typography.
Map video frames onto 3D space using transform: rotateY(deg) translateZ(px) within a perspective container.
Real-time message bubbles animating upward with highlight pills appearing over entity tokens.
Use interpolate on scroll offsets with spring-staggered chat bubble opacity and translateY.
Purpose. Establish paradigm shift regarding machine hearing.
Execution. Crisp centered text over black screen.
Purpose. Illustrate acoustic echo cancellation failure in modern voice agents.
Execution. Smartphone voice UI tripping over its own audio loop.
Purpose. Introduce the fix with Soniox STT v6 branding.
Execution. Bold minimalist typographic card.
Purpose. Demonstrate seamless agent-user conversation and list capabilities.
Execution. Floating chat UI and scrolling feature sheet over winter aerials.