SDF Viewer

Introduction
Features / future plans
Demo
Workflow
Integrations
Implementation details
More demos

Introduction

A fast and cross-platform Signed Distance Function (SDF) viewer, easily integrated with your SDF library.

A Signed Distance Function/Field (SDF) is an alternative approach to design 3D objects. In its simplest form, it is just a function that maps each point in 3D space to the signed distance to the closest surface of the object. It can underestimate the distance to the surface, but it can never overestimate the distance. A negative distance means that the point is inside the object.

SDF libraries will provide you with a wide range of SDF primitives, both 2D (rectangle, circle, text, etc.) and 3D (box, sphere, cylinder, triangle mesh, etc.), and operators (union, difference, intersection, extrude, revolve, etc.) to use in your designs. This application will take your code (written in any language) and quickly and interactively render it to a window. The main objetive of this app is to speed up the design/modeling phase by rendering the changes to your SDF as fast as possible.

I use this to design objects for 3D printing, but it can be used for any 3D modeling task. If you are looking for inspiration or want to learn how to build your own SDF library, check out shadertoy.

Features / future plans

Demo

👉 Click to run the latest web demo 👈

This (outdated) example loads and renders an SDF with custom materials at the maximum framerate of 60 FPS on an integrated graphics card (i7-9750H). The initial load of the SDF is also interactive at 30 FPS (configurable): it quickly loads a low-resolution version of the object and iteratively increases the level of detail until the SDF is fully loaded. Parameters are configured from the UI, also rendering the changes in real time.

Another slightly more complex example (see sdf-viewer-go for more information):

It renders an object defined using the Go language.
It recompiles and reloads the SDF when a change in the code is detected

Workflow

These are the steps to follow to start using SDF Viewer.

Choose an integration depending on your preferred language and SDF library.
Copy an example project from the integration's repository.
Compile the source code to generate a WebAssembly file (see documentation of repo).
Display the WebAssembly file that defines the SDF with the app.
1. If running the native app, use the CLI to set the path to the file or URL (sdf-viewer app url <wasm-file>).
2. If running on a web browser, start a server to provide the file and add ?cliurl=<link-to-wasm-file> to the URL.
3. In both cases, you can also use the Settings window inside the app to set the URL.

Any change to the sources of the SDF would require you to repeat steps 3 and 4 to display the updated version. This is a bit cumbersome, so the server subcommand was created to automate these steps. You give it a set of files or folders to watch, a compile command and the generated wasm file path and it will automatically perform these steps for you. It will also serve the wasm file at an URL that you can give the app and, in addition, it will notify the app of any update, automatically providing the new wasm file.

The server subcommand simplifies the workflow to:

Start the server subcommand with the correct arguments (see server --help or UI menu bar).
Start the app subcommand pointing to the server's URL (see app --help or UI menu bar).
Profit! When you modify and save your source code, the new SDF will automatically be displayed by the app.

Note that all subcommands are also available in the menu bar of the application window. You can read the docs and execute them from the app itself!

Exporting

Once you are ready to export your SDF, you can use the mesh subcommand (or UI button) to export it as a standard triangle mesh. You'll have to select and configure a meshing algorithm. The output is in the PLY format, as it is simple (text-based), and can contain material information embedded in the same file. You can easily view it and convert it to other formats with tools like meshlab. You can also use Blender to perform a Smart UV Project, bake the vertex colors to a texture and then simplify the mesh without losing the colors.

Note that exporting a triangle mesh is a lossy operation (the triangles of the mesh only approximate the underlying SDF), and you should keep the source code or the wasm file in order to export higher quality meshes in the future.

Integrations

Repository	Language	Libraries	Features	Notes
sdf-viewer	Rust	Pure Rust	Core Hierarchy Parameters	Used for the demo / feature showcase Available as a library (`sdfffi` feature, usage)
sdf-viewer-go	Go	Pure Go SDFX SDF	Core Hierarchy Parameters	May be used as a guide for implementing your own integration due to the simplicity of the Go language

Feel free to create integrations for other languages and frameworks and add them to this list!

It is very simple to integrate with other languages and frameworks. The only requirement to provide support for a language is the ability to compile to WASM exporting at least the core methods (getting bounding box and sampling the distance at any point).

Implementation details

Rendering

The renderer is a GPU-accelerated raytracer. To take the SDF definition written for the CPU and render it with the GPU, I had to fill a 3D texture that is then raytraced by a shader. This 3D texture represents samples of the SDF in the 3D grid that contains the object. Each sample contains the distance to the surface and some other material properties like the color and roughness.

Afterward, this GLSL shader does the actual rendering. This shader is applied to a cuboid mesh that represents the bounding box of the object. The mesh is useful for only raytracing the part of the screen that may reach the object, and for extracting the rays for each pixel from the hit points. The shader simply walks along the ray for each pixel, moving by the amount of distance reported by the SDF on each position. If the surface is reached at some point, the normal is computed and the lighting is applied for the material saved in the closest voxel. To get the distance at a point that does not match the grid, interpolation is applied, leading to round corners if the level of detail is not high enough.

The distance function must always be equal to or underestimate the distance to the closest point on the surface of the 3D model. An invalid SDF would cause rendering issues such as "stairs" when looking at a flat surface at an angle. Using this renderer is a nice way of testing for issues while developing objects for an SDF library.

I chose raytracing instead of meshing as building a detailed mesh is slower. While loading the SDF into the 3D texture mentioned above, the shader is capable of rendering a real-time preview of the object, which provides much better interactivity. This is enhanced by the fact that I do several passes to the grid slowly increasing the level of detail by filling more voxels with data, in a way similar to bitmap interlacing.

A high-quality meshing algorithm that preserves sharp features should be applied to get the final model, but the objective of this app is to interactively render previews while designing 3D models through code.

Building

All releases include builds for most platforms. Follow the release.yml workflow to learn how to build the project by yourself.

More demos

Running on Android

After using Termux (a terminal app for Android) to install all dependencies and compile from sources SDF Viewer, SDF Viewer Go and TinyGo, you can run everything on your phone without the need for any other device. This demo follows the commands from SDF Viewer Go.

You can even do everything on the same screen:

ksindi/sdf-viewer