vrm2sl

March 31, 2026 · View on GitHub

screenshot

VRM to SecondLife Avatar bone converter.

Usage

cargo run --manifest-path backend/Cargo.toml --bin vrm2sl -- input.vrm output.glb

or after install/build:

cargo install --path backend --bin vrm2sl
vrm2sl input.vrm output.glb

Important

Do not RESELL or REDISTRIBUTE avatars that include copyrighted assets (such as those sold on Booth) on marketplaces. In most cases, use other than for personal use is prohibited.

Analyze only (no export)

vrm2sl input.vrm output.glb --analyze-only --report report.json

When analysis or conversion completes, a desktop notification is sent (on macOS, via osascript).

Main options

  • --target-height <cm>: target avatar height (default 200)
  • --manual-scale <n>: extra scale multiplier (default 1.0)
  • --resize on|off: texture auto-resize policy (default on)
  • --resize-method bilinear|nearest|bicubic|gaussian|lanczos3
  • --load-settings <project.json>: load project settings
  • --save-settings <project.json>: save current project settings
  • --report <report.json>: save analysis report JSON

Tauri IPC integration layer

The Rust side exposes IPC-friendly functions designed for easy integration with Tauri invoke.

  • ipc::analyze_vrm_ipc
  • ipc::convert_vrm_to_gdb_ipc
  • ipc::save_project_settings_ipc
  • ipc::load_project_settings_ipc

Each function accepts String paths and serializable request structs, with input/output shapes that can be called directly from the UI layer.

Tauri + Vuetify app (integrated)

tauri-vuetify-starter is integrated into this repository.

  • Frontend: frontend/
  • Tauri backend: backend/
  • Rust core implementation: backend/src/

Desktop app run

pnpm install
pnpm dev:tauri

From MainContent in the UI, you can:

  • Select a VRM file
  • Run analysis (bones/vertices/textures/upload cost estimate)
  • Save/load settings JSON
  • Export .glb

Desktop notifications are sent when analysis/conversion completes.

Implemented (core MVP baseline)

  • VRM/GLB input loading
  • VRoid model validation (non-VRM/unsupported source is rejected)
  • Required humanoid bone presence check
  • Required humanoid parent-child hierarchy validation
  • VRM → Second Life bone-name mapping for core humanoid bones
  • Uniform avatar scaling toward SL default height (200cm)
  • Mesh statistics (vertices/polygons) and 65535 vertex-limit diagnostics
  • Texture resolution diagnostics and upload-fee estimate (before/after resize)
  • Removal of VRM extension references/extras from output JSON chunk
  • Removal of animation and morph target entries
  • Project settings JSON load/save API
  • Output .glb (GLB container) generation

Notes

  • Current implementation focuses on Rust core pipeline and CLI.
  • Texture auto-resize currently affects validation/estimation and option handling; embedded image payload rewrite is not yet enabled.
  • Full hierarchy reconstruction, inverse-bind full regeneration/writeback, and advanced UI/preview workflow are planned next steps.

Stability Notes (Eyes/Face)

Recent fixes for face/eye instability (cross-eye, missing iris, flicker) are now part of the default pipeline.

  • Conversion side:
    • skin.skeleton prefers mPelvis when available.
    • Face skin keeps mHead/mEyeLeft/mEyeRight joints.
    • Hair-like tiny secondary skins may be simplified to mHead for stability.
    • Bind-pose correction excludes tiny face eye skins to avoid eye drift.
  • Preview side (frontend/src/components/VrmPreview.vue):
    • Eye materials are handled separately from lash/brow materials.
    • Iris/highlight use anti-z-fighting settings (polygonOffset) to reduce depth sorting artifacts.
    • Upper-body BVH retargeting is enabled (wrists are still filtered to reduce hand collapse).

If eye placement looks wrong, validate in this order:

  1. Regenerate output with current backend (cargo run --manifest-path backend/Cargo.toml --bin vrm2sl -- <input.vrm> <output.glb>).
  2. Confirm skin topology with python3 vrm/inspect_output.py.
  3. Check face skin influence with python3 vrm/inspect_skin0_weights.py.
  4. Compare eye-weighted centers with python3 vrm/inspect_eye_vertex_positions.py vrm/output.glb.

Animation Attribution

Contains animation data © Linden Research, Inc.
Licensed under CC BY 3.0
https://creativecommons.org/licenses/by/3.0/

Modified for use in this tool.

Frontend-side asset note: see frontend/README.md.

Documentation

See Contributing for development and coding conventions.

Testing

  • Run tests with cargo test.
  • Name unit tests in Given/When/Then style for behavior clarity.
  • Example pattern: given_condition_when_action_then_expected_result.

License

©2026 by Logue. Licensed under the MIT License.

🎨 Support This Project

vrm2sl is an open-source tool for bridging the VRM avatar ecosystem and Second Life, combining a Rust core pipeline with a Tauri + Vue 3 desktop frontend. Developing and maintaining it requires continuous research into bone mapping, texture policy, and SL-specific quirks.

If this tool saves you time or helps bring your avatar to Second Life, consider sponsoring to support ongoing development.

GitHub Sponsors