Summary

I would like to propose providing Avatar Shaders with limited access to avatar deformation and spatial relationship data.

This could include:

Pre-skinned vertex data

Bone matrices and bone transforms

Limited avatar spatial relationship queries (such as distance-to-surface or similar performance-safe alternatives)

The following restrictions should apply:

Read-only access only

Accessible only for the avatar currently being rendered

No access to other players' avatar data

No raw mesh export functionality

No mesh modification APIs

No runtime scripting requirements

No GPU buffer write access

The goal is to enable advanced shader-based anti-clipping and corrective deformation systems while maintaining VRChat's existing security, performance, and platform compatibility standards.

Motivation

Clipping between the body and clothing remains one of the most common and difficult problems in avatar creation.

This is especially noticeable during:

Full Body Tracking

Dance and MMD animations

Sitting and crouching poses

Extreme arm and leg movement

PhysBone-driven secondary motion

Current solutions such as body hiding, stencil masking, corrective blendshapes, and manual mesh edits can reduce clipping, but often require significant manual work and may introduce visual artifacts.

Transparent clothing is particularly challenging because hiding clipped body parts can create visible holes that often appear more unnatural than the clipping itself.

Some creators have already demonstrated shader-driven anti-clipping and deformation techniques using depth-based approaches. These projects are fascinating and show the potential of advanced corrective systems.

However, such solutions are often complex to author, view-dependent, difficult to maintain, and limited to specific use cases. Seeing both the potential and the limitations of these approaches is what inspired this proposal.

Predictive Anti-Clipping and Corrective Deformation

Access to pre-skinned vertex data and bone matrices would allow creators to develop predictive deformation systems that anticipate clipping before it becomes visible.

Potential applications include:

Dynamic anti-clipping deformation

Clothing expansion around high-risk clipping areas

Pose-aware corrective deformation

Transparent clothing correction

Volume preservation systems

Dual Quaternion Skinning (DQS)

Optimized Centers of Rotation (OCoR)

Pose Space Deformation (PSD)

For example, clothing could dynamically expand around the hips, shoulders, underarms, or other deformation-heavy regions during extreme poses rather than relying entirely on body hiding or manually authored corrective shapes.

Reactive Anti-Clipping Systems

While predictive systems can significantly reduce clipping, some situations are difficult or impossible to anticipate reliably.

Examples include:

PhysBone-driven motion

Dynamic accessories

Secondary motion systems

Transparent garments

Complex layered clothing

To address these situations, limited avatar spatial relationship queries could enable reactive anti-clipping systems that respond to intersections after they occur.

Possible implementations could include:

Distance-to-surface queries

Signed Distance Field (SDF) sampling

Surface proximity information

Other performance-safe spatial relationship APIs

Importantly, this proposal does not require exposing raw mesh data, unrestricted vertex access, or cross-avatar information.

The exact implementation could be determined by VRChat based on performance, security, Quest compatibility, and future maintainability considerations.

Benefits

This proposal does not request VRChat to implement any specific deformation technology.

Instead, it would provide creators with the foundational data necessary to develop their own solutions entirely through shaders.

Pre-skinned vertex data and bone information would enable predictive corrective systems, while limited spatial relationship queries could enable reactive anti-clipping systems that respond to actual intersections.

Together, these capabilities could unlock a new generation of avatar quality improvements, reduce clipping artifacts, improve transparent clothing workflows, and encourage innovation within the existing Avatar SDK framework while preserving VRChat's current security and performance model.