1 unstable release

0.1.0 Feb 15, 2024

#145 in Rendering

22 downloads per month

BSD-2-Clause

140KB
4K SLoC

Inox2D

Officially supported experimental Rust port of Inochi2D.   Discord

 

The Inox2D workgroup provides support in the #inox2d channel on the Inochi2D Discord.

Currently this library and the specification is in a prototype state, it is not recommended to use this library in production.

 

Rigging

If you're a model rigger you may want to check out Inochi Creator, the official Inochi2D rigging app in development.
This repository is purely for developers and is not useful if you're an end user.

 

Status

Both INP and INX files get parsed correctly.

Both renderers (OpenGL, WGPU) now work on all models we could test them on (Aka, Midori, Arch-chan). The newer models which use the MeshGroup feature don't work yet though.

Support for mesh groups and animations is on the way!

Feature tree

  • Parsing
    • INP format
    • INX format
  • Rendering
    • OpenGL
      • WASM (WebGL)
    • WGPU
      • WASM (WebGL)
    • Draw List
  • Parameters
    • Deforms (mesh vertex offsets)
    • Values (node transform offsets)
    • Z-sort
  • Physics
  • Animations
  • Mesh groups

INP and INX parsing

cargo run -p inox2d --features owo --example parse-inp path/to/puppet.inp

Parsed foxgirl

OpenGL renderer

cargo run -p render-opengl path/to/puppet.inp

OpenGL-rendered Arch-chan

WebGL demo

See the render_webgl example.

WebGL-rendered Aka

WGPU renderer

cargo run -p render-wgpu path/to/puppet.inp

WGPU-rendered Arch-chan

 

Implementation

Inox2D aims to support all features currently present in the standard D implementation.

Inox2D is designed to be extensible. Nodes are extensible through a generic InoxData<T> enum which has a Custom(T) variant. Every other part of the library accounts for it: the OpenGL renderer accepts any struct that implements the CustomRenderer trait to be able to render your custom nodes, and the deserialization functions accept generic Fns for deserialization of custom nodes when it is relevant.

 

Optimization on OpenGL

Implementation language OpenGL calls
Inochi2D reference* D 3076
Link Mauve's inochi2d Rust 551
Inox2D Rust 1639

The OpenGL renderer on Inox2D has a few simple optimizations that result in fewer OpenGL calls:

  • it uses a simple OpenGL cache to avoid making calls when the resulting state won't change,
  • it only uploads a model's part textures once instead of every frame.

* Reference implementation is subject to change as optimisation passes are done, additionally code is more geared towards readability than performance for implementers to be able to more easily use it as reference.

 

License

This project is licensed under the 2-Clause BSD license. See LICENSE for details.

Dependencies

~18MB
~232K SLoC