Type a topic. Get the recap.
An agentic recap build. Drop in a category, a theme, or a bare phrase — an agent grasps the trending story at its centre, runs the recap workflow, and renders a vertical video, an analytics brief, a 3D sources-pull radar, a vertical relationship map and the cast behind the story. The story stays centralised on its originator.
Give it a category. Watch the agent work.
Type a category, a theme, or direct language. The loading screen streams every step; the result renders below in the live format.
What the gathered sources add up to
Vertical recap & its relationship map
Buffering slowly? Open the recap video in a new tab.
9:16 relationship map — single-spine vertical chain, top to bottom, matching the video aspect.
Trend summary, sources & cast
highlights
where the sources pull · 3D radar
Each side's polygon is the pull its sources exert across the debate's dimensions. Drag to spin it; move your cursor over it to tilt.
legend & sources
cast · protagonists & antagonists
The principals the story orbits — each with a retrieved image (team crest, headshot or brand mark) that falls back to a monogram if none resolves.
story points
The recap, as a GIF.
Share-ready, rounded, infinite loop.
An agent that keeps the story central.
Grasps the topic
Turns a category or phrase into a concrete trending story, grounded in real TikTok search signals.
Tiers every clip
Scores sources into main / updates / commentary / tertiary and drops everything off-topic.
Analytics + 3D radar
A metrics card plus a movable 3D sources-pull radar across the debate's dimensions.
Cast with images
Protagonists and antagonists with retrieved crests/headshots and fold-in lines.
Chinchilla build vs. Higgsfield AI models.
Our builds are deterministic code + real source clips. Higgsfield’s lineup is generative AI — different tools, different jobs.
Forward Recap build
The full agent + tiered pipeline, the live result format (analytics, 3D radar, cast images), and the loading-screen UI — as a downloadable .md for AI agents.