Compares voice AI models on naturalness, performance, and cost
The Hook
“Good morning, Theodore! It's sunny, 22 degrees. You've got a meeting at 10, so let's get you up. / Wake me at 7:30. / Of course! Alarm set for 7:30. / No, no, no, no, wait! / Yep. We've all been there.”
Product Reveal
“And soon, voice is going to be everywhere. So it has to do better. Introducing Cekura Bench. To put leading voice AI models to the test in real conversations. Cause...”
Feature Teaser
“It has to understand what you're asking, no matter where your accent is from. / Book me a shoe cleaning next week? After five. / Keep up when you change your mind. Because let's be honest, we all do. / Can you please add 2 oranges and a box of strawberries... Careful, honey! Actually, make that 4 oranges. / Hear you clearly, even when everything around you gets loud. / Bay three, the one on your left.”
Call to Action
“Every voice AI makes these promises. So we benchmarked every leading model, and every category crowned a different winner. Not to call anyone out, but to help you get there. Cekura Bench. Build the voice that leads the future.”
The Hook
“(No spoken dialogue — alarm chimes softly as Theodore stares at screen in disbelief)”
Synthesizable in Remotion · key components of the ai-benchmark-showcase template.
A central glowing spherical particle orb pulsing and shedding luminous data specks.
Use spring() to animate scale of a radial-gradient div while distributing animated absolute divs via Math.cos and Math.sin.
Cards stagger slide upwards with scaling crown badges and typography reveals.
Map card items with frame delay using spring({ frame: frame - index * 6 }) driving translateY and opacity.Purpose. Establish universal frustration with current voice assistant failures.
Execution. Comedic missed alarm waking sequence ending in a freeze frame.
Purpose. Pivot from problem to vision for reliable voice AI.
Execution. Abstract eclipse and branding animation.
Purpose. Demonstrate accents, real-time corrections, and noisy environments.
Execution. Three cinematic vignettes with live-updating smart UI.
Purpose. Present unbiased empirical evaluation results.
Execution. Animated metrics grid ranking leading speech models.
Purpose. Reinforce brand mission and deliver final comedic payoff.
Execution. Clean brand lockup followed by evening alarm callback.