More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIMuseMuseDotsDots+21 more

    Kivo White-Model Video

    1

    Use this skill in Codex, Claude Code, or another SKILL.md-compatible agent: type a shot prompt, and the agent builds a local white-model previs — scene, camera, preview, and MP4 — before any optional Kivo upload.

    Free

    Security scanned
    Kivo White-Model Video

    Kivo White-Model Video

    Example session with this skill installed

    Create a 6-second greybox previs shot of a detective walking down a rainy alley, stopping under a streetlight and turning as the camera orbits. Give me the beat table, a readable night-lit scene, and a locally rendered 1280x720 24fps MP4 reference.

    • Read your context and instructions
    • Compiled the kivo white-model video

    This is an agent skill, not a standalone app. Install the folder that contains SKILL.md into your agent's skills directory, then ask the agent for a previs in plain language.

    Codex (see Build skills): copy the package to ~/.agents/skills/kivo-white-model-video/ for all your projects, or to .agents/skills/kivo-white-model-video/ in a repository. Mention it with $ or /skills, or just describe the shot.

    Claude Code (see Extend Claude with skills): copy it to ~/.claude/skills/kivo-white-model-video/ (personal) or .claude/skills/kivo-white-model-video/ (project). Ask Claude to use it, or run /kivo-white-model-video.

    Cursor (see Agent Skills): place the same folder in .agents/skills/ or .cursor/skills/ (project), or ~/.agents/skills/ or ~/.cursor/skills/ (user), and optionally invoke it with / in Agent chat.

    Example prompt

    Use the kivo-white-model-video skill: a detective walks down a rainy alley at night, stops under a streetlight, and turns as the camera orbits around him.

    The agent should return a beat table (time, action, camera), a local preview URL, and — when you asked for a video — an MP4 path with duration and a checksum. You can then request revisions on the same shot.

    The first pass is a single 1280×720, 24 fps clay/greybox shot, typically 5–8 seconds. The agent needs Node.js 22+, Chrome or Chromium with WebGL, and FFmpeg/FFprobe on the machine that renders. Local previs does not require a Kivo account. Connecting Kivo Studio MCP is optional, and only for uploading an approved file to a project library or starting Seedance generation after you authorize the project, confirm rights to the file, and approve any spend.

    This is independent community packaging of the workflow, not official documentation or a quality guarantee. Homepage: kivst.com.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    An agent skill for previs before AI video generation. Give it a single prompt (e.g. "a detective walks down a rainy alley, stops under a streetlight and turns as the camera orbits") and it writes a beat table, builds the scene from reusable helpers (mannequin characters with walk/run/turn/sit/hug, alley/park/office sets, rain, readable night lighting), animates camera rigs (track, true orbit, push-in, crane-up and pull-back), self-checks frames at each beat, and renders a 1280x720 24fps MP4 locally with HyperFrames + Three.js. Optionally uploads the reference to Kivo Studio via its MCP for Seedance-based generation. Tested on three prompts with passing renders.

    Requires local Node.js, Chromium, and FFmpeg for rendering. Motion must be non-accumulated and seekable by timestamp. Requires Kivo MCP connector for library uploads.

    How to use

    1. 1

      Connect your agent to the Agensi MCP

      Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup

    2. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 3

      Ask for what you need

      Your agent loads this skill and uses it on your task. It always asks before any payment.

    Reviews

    No reviews yet

    Be one of the first to try it. Every listed skill goes through the trust checks below.

    Security scanned

    Checked by our 8-point scan before listing

    Fresh listing

    Recently published to Agensi

    Free forever

    No account required to browse

    Trust & safety

    Security scan

    Security scan passed 5 days ago

    • Free with an Agensi account

    Requires

    hyperframes
    Node.js 22+
    Chromium with WebGL
    FFmpeg/FFprobe
    Node.js
    macOS

    Permissions

    Terminal / Shell
    Browser
    Read Files
    Network Access
    Write Files
    Listed5 days ago

    What's inside

    Frequently Asked Questions

    This skill eliminates "hallucinated" motion by providing a deterministic 3D white-model reference. It ensures character blocking, camera paths, and timing are locked in before you spend credits on final AI video generation.

    The skill requires a Node.js environment with Chromium and FFmpeg installed for rendering. It is designed to work with AI agents that support MCP (Model Context Protocol) and is optimized for the Kivo Studio ecosystem.

    Yes, the skill generates editable Three.js code in a dedicated workspace. You can manually adjust geometry, motion paths, and lighting recipes within `scene.mjs` to achieve the exact blocking required for your shot.

    You receive the full skill package, including the scene scaffolders, character mannequin system, camera orchestration logic, and the Kivo MCP connector for cloud asset management.

    This skill creates 24fps MP4 files that are specifically formatted as "Seedance" motion references. This allows you to guide the motion of high-fidelity models with frame-accurate precision.

    The skill uses a seekable timestamp system rather than frame accumulation. This means you can jump to any point in the timeline to check a pose or camera angle without rendering the entire sequence from the start.

    New in Image & Video