Brioche Puppet

I am currently working on the free and open-source VTuber software Brioche Puppet.

As a rendering engineer, the rendering is implemented by myself. But the third-party libraries JoltPhysics and MeidaPipe are used for ragdoll physics and motion capture.

1. Asset Import

Brioche Asset Import

1-1. Mesh and Animation

• MMD (PMX, VMD)
• glTF (glTF, glB, VRM, VRMA)

1-2. Image

1-2-1. Albedo Image

• WebP
• PNG
• JPEG

1-2-2. Illuminant Image

• OpenEXR

1-3. Environment Map (Environment Lighting Image)

• Equirectangular (Latitude-Longitude) Map
• Octahedral Map
• Cube Map

2. Motion

Brioche Motion

• Motion Capture
  • Video Capture
  • Motion Detector
• Animation
  • IK (Inverse Kinematics)
    • Reaching IK (Target Position)
      • Two Joints IK
      • Three Joints IK
      • CCD (Cyclic Coordinate Descent) IK
      • FABRIK (Forward And Backward Reaching Inverse Kinematics)
      • Biped Foot IK
      • Quadruped Foot IK
    • Aim IK (Target Direction)
      • Look At IK
  • Skeleton Mapper
    • Ragdoll Mapping
    • Animation Retargeting Mapping
  • Ragdoll
    • Ragdoll Kinematics Controller
    • Ragdoll Motors Controller

3. Rendering

Brioche ANARI

3-1. Rasterization

• PBR (Microfacet Model: Trowbridge Reitz)
• Toon Shading
• SSS (Skin: Subsurface Scattering)
• LTC (Area Lighting: Linearly Transformed Cosine)
• IBL (Environment Lighting: Spherical Harmonics + Split Sum Approximation)
• VXGI (Global Illumination: Clipmap Cone Tracing)

3-2. Ray Tracing

• ReSITR (Reservoir-Based Spatiotemporal Importance Resampling)
• SVGF (Denoiser: Spatiotemporal Variance Guided Filtering)
• SSS (Skin: Subsurface Scattering)
• SHaRC (Spatially Hashed Radiance Cache)

3-3. Nerual Rendering

• NS (Nerual Shading)
• NeRF (Neural Radiance Fields)
• NRC (Neural Radiance Cache)