The Realtime AI Company.
The Hook
“Voice AI is the new default user interface, but what matters to people isn't how it sounds, it's how it makes them feel. We built real-time AI for consumer-facing applications because we want everyone to experience the magic of voice AI.”
Product Reveal
“Hey, I'm Igor, Chief Science Officer at Inworld, and today I want to introduce you to our best TTS model family yet, the Inworld TTS-2. We've rebuilt the entire serving stack, and the model's architecture in order to support ultra-low latency in real-time production applications. TTS-2 is the most versatile speech synthesis model out there. Similarly to large-language models, it's now promptable, and understands the conversational context. It hears what the users say.”
Feature Teaser
“A good voice actor knows how to adapt to the moment. With voice steering, you can instruct the model in natural language, as you would an actor. Cross-lingual support allows dynamic switching between over 500 dialects, with the same speaker identity. And it all happens under 150 milliseconds to feel organic and real-time.”
Feature Teaser
“Hey Prashanth, can you come and try this out? Are you excited for the launch? [in an excited tone] Of course! I'm very excited to launch TTS-2 today! Impressive. Anastasiya, you want to check this out? Okay, let's see. Are there many bugs left in you? [in a sarcastic tone] After your checks? Not a single one survived! That's good. Hey Andreas, can you come and try it? Yo. I'm having a rough day, do you have any advice? [in a calm and reassuring tone] Calm down, breathe, everything is going to be okay. That was crazy! We should make ads in all languages now!”
Problem Agitation
“Hi, my name's David. I'm cofounder and CTO of Livekit, and Livekit is a platform that helps developers to build voice agents. I was really impressed with how steerable the model is. The ability to prompt the voice to sound different given the same text, but having a full range of different types of emotions it could display is really impressive. It really felt like talking to a person. If you're using another TTS model and have not checked out Inworld yet, I would highly recommend giving them a look. They went from launching their first TTS model, to topping artificial analysis charts in just about a year.”
Call to Action
“TTS-2 is live today, and you've probably already spoken with it along with hundreds of millions of users. Try it today at realtime.ai.”
Synthesizable in Remotion · key components of the executive-announcement-live-demo template.
A circular particle sphere undulating dynamically in response to speech frequency and amplitude.
Generate a Canvas element in Remotion; compute simplex noise offsets modulated by audio amplitude extracted via useAudioData.
Semi-transparent rounded badges floating on screen with spring-based entrance and subtle sinusoidal vertical drift.
Use spring() from remotion for entrance stagger, map frame count with Math.sin() for subtle floating position.
Centered monochromatic logo mark with clean serif/sans typography fading in with slight upward translation.
Interpolate opacity from 0 to 1 and translateY from 12px to 0px using Easing.out(Easing.cubic) over 25 frames.
Purpose. Establish voice AI as the primary future interface.
Execution. Walking executive tracked into office environment.
Purpose. Unveil TTS-2 architecture and sub-150ms performance.
Execution. Chief Science Officer direct address with title card.
Purpose. Demonstrate emotion steering and cross-lingual fluency across Hindi, Russian, Spanish, and Mandarin.
Execution. Casual desk interactions with colleagues reacting to live UI.
Purpose. Provide trusted third-party benchmark proof via LiveKit CTO.
Execution. Professional boardroom documentary interview.
Purpose. Showcase ecosystem traction and direct users to live trial.
Execution. Floating partner logo badges followed by minimalist end card.