Turn complex documents into AI-ready data
The Hook
“70 million pages for 2.5 million users. That's how much we process while building YouLearn.”
Problem Agitation
“We had to handle 500-page textbooks, handwritten notes, and scanned documents with broken tables, which is why we built our own document parser...”
Product Reveal
“...and launched it as an API.”
Problem Agitation
“Once we shared it, we didn't expect teams from healthcare, government, and insurance to be the first to reach out. But after talking with them, we realized why they were still looking for a solution.”
Problem Agitation
“Either they built their own pipeline and as they scaled, maintaining it became a constant burden, or they were using existing providers and had to deal with trade-offs between accuracy, cost, and speed that wasn't in their control.”
Product Reveal
“So we built Hanji. You send us your documents, and we tune your API to your workflow with the accuracy, cost, and speed you need.”
Feature Teaser
“And even before tuning, Hanji holds the highest accuracy on our benchmark.”
Feature Teaser
“And you can try it yourself. Upload a patient intake form and extract the handwritten values, pull fields from a degraded government scan, or recover structured data from dense tables.”
Feature Teaser
“After we tune it to your docs, the results improve even more.”
Feature Teaser
“For one YC healthcare company, we improved their text accuracy from 71% on the first pass to 96.5% within 5 days. For a second team, we reduced their end-to-end latency from 90 seconds to under 1 minute. And for a third, we cut processing cost by 75% by adding a routing step.”
Call to Action
“So try Hanji on our playground today, then book a 15-minute call to see how we can tune it to your needs.”
Call to Action
“(No spoken dialogue — subtle outro chime)”
Synthesizable in Remotion · key components of the founder-scale-saas-launch template.
Displays numeric counter animating upwards while nested status bars transition from red misread states to green verified states.
Use interpolate() on frame to drive displayed integer and CSS width, pairing with spring() scale on status pill transitions.
Simulates document region selection with colored semi-transparent boxes that resolve into structured key-value sidebar elements.
Animate SVG rect opacity and border dashoffset synchronized with staggered slide-in of key-value text rows.
Progress bars grow left-to-right with trailing standard deviation error bars and bold percentage labels.
Drive flex item width via spring({ damping: 15 }), chaining delayed opacity reveals for error bars.Purpose. Establish instant founder credibility through high processing volume.
Execution. Founder on camera paired with oversized bold metrics.
Purpose. Illustrate enterprise pain points with in-house pipelines vs static vendor trade-offs.
Execution. Interactive two-column comparison card with live error counters and locked controls.
Purpose. Demonstrate tangible ROI across accuracy, latency, and operational expense.
Execution. Three-part tabbed metric showcase featuring before-and-after green delta tags.