Content Creation / Internal

Video Overlays

Engine to automatically generate video overlays (9:16, alpha .mov) from speech scripts — AI decides highlighting points, and the framework renders.

Overview

An engine generating overlay animations for talking-head videos: inputs scripts, AI computes highlights and overlay types, generating transparent overlays for CapCut.

Created for our own video production pipeline — making talking videos more engaging with automated graphic elements.

Output includes an HTML showcase file displaying overlay placements and download links for the editor.

The problem to solve

Manually creating overlays for each video is time-consuming (~10 effects take hours). Needed automation to focus on recording and speaking.

How we approached it

Script input → AI only decides where and what type of overlay to use (minimal tokens); rendering is handled by the framework, costing almost $0.
15 pre-built overlay templates, auto-calculating duration based on Vietnamese speech rate.
Exports alpha .mov files (ProRes 4444) optimized for CapCut, with HTML showcase for editors.
Agent-agnostic: works with anti, cursor, claude, etc. — just convert skill to your tool.

What the product does

Auto Highlight Detection

AI reads script, suggests overlay segments and types.

15 Overlay Templates

Title/Hook, Text highlight, Branches, Stats, Lists, Bullets, Chips, Bars, Donuts, Quotes, CTAs...

Auto Duration Calculation

Based on syllables and Vietnamese speech rate, matching voiceovers.

Alpha .mov Export (ProRes 4444)

Drop directly into CapCut or similar editors.

HTML Showcase Output

HTML output showing which overlay goes where, with download links.

Quick Configuration

Fast setup of logo, info ~3 minutes; includes script examples for testing.

Technical challenge

Problem

Render beautiful and cheap video overlays without relying on expensive AI video generation APIs.

Solution

Use AI only for decision-making (low tokens), delegating rendering to Remotion + animation libraries → high-quality transparent .mov export for ~$0.

Results

Reduced overlay creation time from hours (~10 effects) to ~5 minutes.
In active use for the video series 'What can the IT industry do?'.
New workflow: record for 20 minutes, wait for generation, upload — done.
Open source on GitHub: https://github.com/tensorlabtech/video-overlays

Highlights

AI decides highlights, framework renders → cost is ~$0
15 overlay templates with auto duration matching Vietnamese speech
Alpha .mov export for CapCut + HTML showcase for editors
Agent-agnostic: runs with anti / cursor / claude (read SETUP README)
Note: NOT full AI video generation — human presence still required

Explore other products

E-commerce / Outsource