Ctrl K
Mindverse Documentation

MindverseField Guide

From the first idea to the next shot: see every step, keep every useful result.

The workspace is a map of decisions

Mindverse is a node-based media workspace. Keep a brief beside a script, a storyboard beside the images derived from it, and the final clips beside their sound. A line between nodes records which source informed the next step.

Brief→
Script→
Storyboard→
Image→
Video→
Edit
  1. Create a project in Workspace and inspect the dotted canvas.
  2. Right-click blank space, add a Text node, then left-click to place it.

Start with the four basic media nodes

Text records and generates instructions; Image creates or edits stills; Video generates motion from compatible input; Audio holds uploaded sounds and generation modes for music and speech. These are real nodes in the current canvas, not marketing-only categories.

  1. Open the blank-canvas right-click menu and locate the Text, Image Generation, Video Generation and Audio entries.
  2. Place one of each, then click them individually and compare their handles and controls before running a paid model.

Use Agent as a collaborator

The Director Agent can interpret an instruction, propose an editable plan and stage a canvas change. A preview is not a finished generation; inspect models, source assets, duration and approval controls before applying.

  1. Open Agent from the left-side canvas bar and describe an achievable single shot.
  2. Review its preview card and choose the relevant Apply or position action only after checking the proposal.

A model is a route with constraints

The codebase connects 302AI and TokenStar text/image routes, HKGAI Seedance and MiniMax video, Kling, Qwen speech, HKGAI music/TTS and OmniHuman. Selecting a label only works when the server has credentials and provider access. Input-port compatibility is more important than the model name alone.

  1. Choose a media node and look at its model dropdown and available input ports.
  2. Read the model-specific chapter before assembling several reference images, audio tracks or clips.

Keep ownership of the result

A selected canvas node, saved project, exported JSON and archived media are different things. A JSON graph preserves structure but is not a complete offline copy of every image, video or voice. Reused portraits and voice recordings require authorization.

  1. After a successful render, inspect the playable output and keep the source asset in a safe location.
  2. If a remote save fails, keep the browser draft, investigate the error and avoid clearing site data.
Note

Provider availability, pricing and media rights are not guaranteed by this guide.

助力大赛,创作继续!

+

今日 2,500 积分已到账。