Skip to content

Desuq Cafe
Documentation menu

Quickstart Preview

Preview. Weft is not yet purchasable, so this page describes what setup feels like rather than giving you copy-paste steps to follow right now. The package ships its own full step-by-step walkthrough, opened from a single menu item inside Unity.

Weft’s own pitch is import to a working grab in about 20 minutes, without writing any solver code. Here is that story.

Import and drop in your avatar

You add Weft to your project through the Unity Package Manager, and its first-party dependencies (Burst, Collections, Mathematics) resolve automatically. Then you drag your rigged avatar into the scene. Any skeleton works, a humanoid FBX, an MMD-style PMX conversion, a VRM import, or a hand-built rig. A humanoid rig unlocks a couple of one-click conveniences later, but nothing about the core setup requires one.

Tell it what should move

Next you mark which parts of the avatar should simulate. A hair chain (a ponytail, a bang, a ribbon) gets a strand component that walks its bones. A full garment with no bone chain, like a skirt or top modeled as a mesh, gets a mesh cloth component instead that binds directly to the renderer. If several strands share the same style, like a skirt’s ring of hem panels, you can group them so they simulate together in one shared world and stay stitched to their neighbors instead of splaying apart.

Keep it out of the body

If a strand or garment could clip through the avatar’s own body, you add a body colliders component. On a humanoid rig this is a single button click: the inspector reads the rig and fits capsules to the hips, torso, arms, and legs in one pass. Every capsule stays fully hand-editable afterward, so the auto-fit is a starting point, not a lock-in.

Paint what can be touched

With the front-end in place, you paint the regions that should respond to a grab directly onto the mesh or strand in Unity’s Scene view, using a brush tool with a live heatmap. Painting by hand on the actual mesh is the intended path, because a script or index-based selection cannot see your avatar’s real shape the way your eyes and hand can.

Add a grabber and a reaction

A grabber component drives the actual grab lifecycle, reading from whatever input source you assign (a VR controller, a mouse, or your own scripted input). Add a second grabber for a second hand, and each tracks its own grab independently, so both hands can hold the same skirt or hair at once. Grabbers also expose hooks you can subscribe to, so a grab can raise a reaction, a blush on a headpat, a small heart on a high five, a giggle on a ruffle. The package ships ready sample hooks you can import and customize.

Hit play and tune by eye

From there it is play mode and feel. A handful of named parameters (how loose or firm the cloth hangs, how snappy the release recovery is, overall energy bleed) let you tune the feel per avatar starting from a safe preset, rather than fighting raw physics constants.

Bringing an existing rig

If your avatar is a converted MMD model, an auto-setup reads its existing rigid bodies, joints, and bones, and proposes the whole chain, collider, and touch zone setup as ordinary components you review and adjust, rather than importing anything blind. If it is a VRM avatar, a separate optional adapter reads its existing spring bone chains and assembles the equivalent Weft setup for you.

Where to go next