reminjp / three-vrm Goto Github PK

It may be fixed by ousttrue/UniGLTF#18

Currently the loader supports only a few properties of Unlit, MToon.

• VRMが提供するシェーダー
• UniVRMで使えるシェーダー

TODO

Unlit

• VRM/UnlitTexture
• VRM/UnlitCutout
• VRM/UnlitTransparent
• VRM/UnlitTransparentZWrite

MToon

Properties

• RenderType: Opaque | Cutout | Transparent | TransparentCutout
• uniform float _Cutoff
• uniform vec4 _Color
• uniform vec4 _ShadeColor
• uniform sampler2D _MainTex
• uniform sampler2D _ShadeTexture
• uniform float _BumpScale
• uniform sampler2D _BumpMap
• uniform float _ReceiveShadowRate
• uniform sampler2D _ReceiveShadowTexture
• uniform float _ShadeShift
• uniform float _ShadeToony
• uniform float _LightColorAttenuation
• uniform sampler2D _SphereAdd
• uniform vec4 _EmissionColor
• uniform sampler2D _EmissionMap
• uniform sampler2D _OutlineWidthTexture
• uniform float _OutlineWidth
• uniform float _OutlineScaledMaxDistance
• uniform vec4 _OutlineColor
• uniform float _OutlineLightingMix
• uniform int _DebugMode
• uniform int _BlendMode
• uniform int _OutlineWidthMode
• uniform int _OutlineColorMode
• uniform int _CullMode
• uniform int _OutlineCullMode
• uniform float _SrcBlend
• uniform float _DstBlend
• uniform int _ZWrite
• uniform int _IsFirstSetup

After implementing VRM.

• .humanoid (Motion data (e.g. BVH. VMD), IK)
• .blendShapeMaster (Morphing)
• .firstPerson (Toggle Object3D.visible, first person camera)
• .secondaryAnimation (Physics)
• Lights

Since VRMloader loads the model asynchronously it might make more sense for it to return a promise?

WIP
https://github.com/rdrgn/three-vrm/blob/master/src/data/VMD.ts

VRM format uses node indices to specify the bone but loaded Scene doesn't hold indices info. So we have to add a new function to get a humanoid's bone by name (string or enum?) or index.

• モデルのボーンマッピング(json.extensions.VRM.humanoid)
• https://threejs.org/docs/#api/objects/Bone
• https://threejs.org/docs/#api/objects/Skeleton

v0.0.10では表示されるモデルがv0.0.11では表示されない不具合があります

モデルデータの読み込みは成功していてscene上には存在しているのですが、可視化されません

確認に使用したモデル: http://3d.nicovideo.jp/works/td38002
(他のモデルも試しましたが同様の現象が起こりました)

v0.0.10

v0.0.11

Windows 10

• ❌ Firefox Developer Edition 62.0b19 (64bit)
• ✔️ Firefox 61.0.2 (64bit)
• ✔️ Google Chrome 68.0.3440.106 (64bit)
• ✔️ Microsoft Edge 42.17134.1.0

iOS 11

• ✔️ Safari (iPhone)

It seems like VRMHumanBoneBoneType is missing some entries, namely:

[
  "leftShoulder",
  "rightShoulder",
  "leftToes",
  "rightToes",
  "jaw",
  "leftToe",
  "rightToe"
]

I am basing myself on this description of the spec.

Publish a ready-to-use package to npm.

In the case where a model is not placed at the origin initially, drag force seems to be applied, even if VRMPhysics.prototype.reset method is invoked after the initial position was set.

The capture below shows the case the model is rotated by (0, π, 0) and translated by (0, 0, -2).

For example, the bangs are floating.