Description
You spot a viral short and want the same thing with your own product, but taking it apart is the hard part: which line goes with which clip, when the captions jump, which beat the cut lands on, all counted frame by frame. hypit lets AI agents such as Claude Code and Codex make video directly: hand it a video and it rebuilds the whole workflow, footage, captions, B-roll and effects included, with everything anchored to words instead of seconds.
That difference matters: change one line of dialogue and the visuals and captions after it shift along, so there is no timeline to re-sync. The same workflow with a new host, product, language or aspect ratio gives you a batch of variants. You do not have to clone anything, either: start from a template, or just describe the video you want and let the agent write it from scratch. Generation models are optional, and visuals driven by front-end code render fully locally with no generation API calls. Install the Skill once, then type /hypit in your agent to get going.
Video cloning:Drop in a reference video and get an editable, re-runnable workflow back, not a script breakdown.
Word-level alignment:Footage, captions and effects hang on the words of the script, so rewriting a line re-flows the timing.
SVML language:Describe a video in HTML-like markup, source on the left and a live preview on the right, easy for an agent to read and write.
Batch variants:Reuse one workflow and regenerate only the parts that change, from one video to a hundred hooks or product versions.
Swappable components:Change the host without touching the captions; use the library, fork it or write your own.
Local parallel rendering:Up to 64 headless Chromium renders at once, with no per-render fee and no added watermark.
Optional models:Avatars, voice-over and B-roll generation can use your own API or local models, billed by the service you choose.
That difference matters: change one line of dialogue and the visuals and captions after it shift along, so there is no timeline to re-sync. The same workflow with a new host, product, language or aspect ratio gives you a batch of variants. You do not have to clone anything, either: start from a template, or just describe the video you want and let the agent write it from scratch. Generation models are optional, and visuals driven by front-end code render fully locally with no generation API calls. Install the Skill once, then type /hypit in your agent to get going.
Features
Video cloning:Drop in a reference video and get an editable, re-runnable workflow back, not a script breakdown.
Word-level alignment:Footage, captions and effects hang on the words of the script, so rewriting a line re-flows the timing.
SVML language:Describe a video in HTML-like markup, source on the left and a live preview on the right, easy for an agent to read and write.
Batch variants:Reuse one workflow and regenerate only the parts that change, from one video to a hundred hooks or product versions.
Swappable components:Change the host without touching the captions; use the library, fork it or write your own.
Local parallel rendering:Up to 64 headless Chromium renders at once, with no per-render fee and no added watermark.
Optional models:Avatars, voice-over and B-roll generation can use your own API or local models, billed by the service you choose.
