video-shortform

Generates 3-10 second short-form MP4 video clips via text-to-video models through a unified media dispatcher.

Updated May 20, 2026
One-click install
npx skills add https://github.com/Keimoshi/Silverloft --skill video-shortform-keimoshi
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: video-shortform
Source: https://github.com/Keimoshi/Silverloft/tree/main/open-design-port/skills/video-shortform
Command: npx skills add https://github.com/Keimoshi/Silverloft --skill video-shortform-keimoshi

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Creating short-form video clips (product reveals, motion teasers, ambient loops) requires knowing each text-to-video model's prompt format, duration caps, and aspect constraints. This Skill standardizes shot planning, prompt composition, and dispatch so a single MP4 lands in the project folder every time. ## Core Features & Use Cases - Model-aware shot planning: Reads project metadata (videoModel, videoLength, videoAspect) and clamps prompts to what Seedance 2, Kling 3/4, Veo 3, or Sora 2 actually support. - Structured shotlist workflow: Plans subject, camera, lighting, motion, and sound before dispatching, so users can redirect cheaply before a slow render. - Unified dispatch: Calls the OpenDesign media contract (od media generate) instead of provider APIs directly, returning one MP4 saved to the project. - Use Case: Ask for a 5-second product reveal of a ceramic mug rotating on a paper backdrop; the Skill plans the shot, binds the project's 16:9 aspect and 5s length to API parameters, and saves the clip for playback in the file viewer. ## Quick Start Generate a 5-second 16:9 product reveal video of a ceramic coffee mug rotating on a soft paper backdrop with warm side lighting.

Frequently Asked Questions about video-shortform

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I generate a short product reveal video with AI?

Plan one shot covering subject, camera, lighting, and motion, then dispatch through the media generate command with the project's model, aspect, and length parameters. The rendered MP4 is saved to the project folder and plays in the file viewer.

Which text-to-video models does this skill support?

It defaults to Seedance 2 but works the same with Kling 3 and 4, Veo 3, and Sora 2. Seedance caps at 10 seconds, Kling 4 supports up to 10 seconds plus image-to-video, and Veo 3 supports 8 seconds with audio.

Can I create multi-shot video timelines with this skill?

No, the skill enforces one shot per turn because current text-to-video models handle single shots well but struggle with multi-cut narratives. Multi-shot timelines belong in a hyperframes or interactive-video skill instead.

What happens when a video generation request fails?

When the underlying model fails due to an NSFW filter, content policy, or timeout, the skill reports the error verbatim without silently retrying. It also never claims a render started unless the media generate command was actually called.

How does hyperframes-html differ from photoreal video models?

hyperframes-html is a local HTML-to-MP4 renderer, not a photoreal text-to-video model. For it, write a concise motion-design brief covering subject, layout, palette, and motion character instead of camera-realism prompts, and dispatch immediately.