It's here! 🥳 Today, we're excited to announce that WebGPU support has officially arrived in the PlayCanvas Editor.

WebGPU on the Rise​

Since its inception back in 2010, PlayCanvas has been layered on top of WebGL. In 2017, we were proud to launch support for WebGL 2.0 in partnership with our friends at Mozilla. April 2023 marked the beginning of a new era for web graphics when Google enabled WebGPU by default in Chrome 113. Since then, WebGPU adoption has exploded and today, Web3D Survey reports that 62.19% of end users can now run WebGPU. With Firefox and Safari due to launch their WebGPU support in the not-too-distant future, expect this number to rise dramatically in 2024.

Why WebGPU Matters​

If you enable WebGPU for your PlayCanvas project, you may not notice much difference to begin with. In fact, we have put a great deal of effort into ensuring your WebGL projects look identical under WebGPU. But over time, there is are a great deal of opportunities to achieve performance improvements due to WebGPU's reduced driver overhead.

Another key feature unique to WebGPU is support for Compute Shaders which allow for general computation on the GPU. Support for Compute Shaders landed in Engine v1.70.0. Here you can see Computer Shaders in action simulating 1 million particles on the GPU:

Click here to run it for yourself in a WebGPU-enabled browser (i.e. Chrome or Edge).

In short, WebGPU represents the future for PlayCanvas and you can expect some incredible advances in performance and functionality over the coming months.

Getting Started with WebGPU​

WebGPU support in PlayCanvas is still considered 'Beta'. There are still some unimplemented features (for example, the run-time lightmapper is still not supported). Therefore, you have to currently 'opt in' to WebGPU support. To do this, open your Project's Settings in the Inspector and expand the RENDERING section. Then update Graphics Devices to include WebGPU (beta).

Once we are satisfied WebGPU support has matured enough, it will become the default.

Your Feedback is Important​

Since WebGPU support is new, we rely on the community for feedback. What works and what doesn't? Please submit an issue if you discover any problems or kick off a new thread on the forum if you want to discuss WebGPU support in more detail. We want to hear what your opinions! 👂

Visual Studio Code is massively popular. In the Stack Overflow 2023 Developer Survey, Visual Studio Code was ranked the most popular developer environment tool among 86,544 respondents, with 73.71% reporting that they use it.

Any PlayCanvas developer building directly on top of the Engine will very likely opt to use VS Code. But if you use the PlayCanvas Editor, you will normally rely on the built-in, browser-based Code Editor. Today, we're excited to give you another option by launching an open-source Visual Studio Code Extension for PlayCanvas.

Benefits​

The PlayCanvas Code Editor is actually built on the Monaco Editor, the beating heart of VS Code. So why use VS Code instead of the PlayCanvas Code Editor?

• GitHub Copilot - leverage AI to help you write PlayCanvas code faster.
• Powerful IntelliSense tools (code completion, parameter info, quick info, and member lists).
• Leverage a huge library of other extensions to accelerate your development.

We ❤️ Open Source​

As you might expect, we have open sourced the VS Code extension under a liberal MIT license!

• VS Code Extension on GitHub

So if you find a bug or have a suggestion, please do log an issue. And for the more adventurous, consider making a code contribution!

We want you to feel empowered to make these tools your own. Let's make them awesome together! 🙌

Get Started Now​

If this all sounds great to you, why not give it a try? Head over to the User Manual for instructions on how to get started:

READ THE DOCS

Today, we are excited to announce the brand new PlayCanvas Developer Site and API Reference Manual!

Developer Site​

Let's start by taking a look at the new Developer Site, home of the PlayCanvas User Manual and Tutorials.

Migrating from Metalsmith to Docusaurus​

Since PlayCanvas was born in 2011, the basic look of the Developer Site has changed little. The content was written in Markdown and converted to a static HTML site using a tool called Metalsmith. Customizing the site to our needs meant we had to write many Metalsmith plugins (for localization, navigation and more) and a lot of HTML and CSS to style the pages as we wanted.

13 years later, the world of static site generators has moved on with lots of exciting new options! We evaluated Docusaurus and decided to migrate over to it.

What's New​

So what made Docusaurus so compelling? First up, the migration was going to be straightforward because, like Metalsmith, Docusaurus consumes Markdown and outputs a static site. And much of the customization we did for Metalsmith is supplied 'out-of-the-box' with Docusaurus. So that's all great. But the new site comes with lots of cool new features we known you're going to love:

🔍 Powerful Search​

The site now integrates Algolia for advanced searching of the documentation.

This is a quantum leap over what came before. Try it (by pressing CTRL + K) - you'll be amazed. 🤯

🌒 Light and Dark Modes​

Easily switch between light and dark themes (by clicking on the sun/moon icon top-right).

🌏 Language Selection​

Switch language from the nav-bar. At the moment, we're shipping with Japanese translations.

👩‍💻 Edit on GitHub​

As you might expect, the Developer Site is Open Source (MIT). Every page now has a handy link to the content on GitHub. Spot a mistake? Now can you fix it yourself and improve the docs for the community!

🧭 Easier Navigation​

At the top of each page, you'll find 'Sidebar Breadcrumbs' which allow you to click back up the page hierarchy.

To the right of each page, you'll find a table of contents which can be a big help, especially for long pages.

And at the bottom of each page, you'll find 'Previous' and 'Next' links that allow you to read through the User Manual sequentially.

API Reference​

We are not just launching a new Developer Site today. We are also launching our brand new, upgraded API Reference Manual!

Migrating from JSDoc to TypeDoc​

Since the start, we have relied on JSDoc to generate our API reference manual. But since then, we have seen the introduction of TypeDoc which offers some major advantages over JSDoc.

🎨 High Quality Default Theme​

While it's not perfect, the default TypeDoc theme is gorgeous and a big improvement over our custom JSDoc theme. It provides:

• Light and dark modes (like the main developer site)
• Links to the source code on GitHub
• Grouping of API into related categories
• Powerful search (press / to activate it)
• ...and much more!

It's a great foundation for us to begin with and we can customize the default theme to our requirements as needed.

🔗 Easy Combining of APIs​

Did you know that PlayCanvas develops frameworks and libraries other than the PlayCanvas Engine? There's PCUI, a front-end framework for web-based tools. There's the PlayCanvas Editor API for automating the interface. And more! The new API reference collects all of the PlayCanvas APIs into a single manual.

Open Source FTW​

Everything we are announcing today is 100% open source.

• Developer Site on GitHub
• API Reference Site on GitHub

We want to empower to community to get involved and make PlayCanvas better for everybody. It's never been easier to submit your first pull request on GitHub. So why wait - get started today! ❤️

Introduction​

3D Gaussian Splatting is a new method for digitizing and rendering real world objects. With gaussian splatting, you can digitize a scene from a few photos using services like Luma Labs or Polycam. These services take the set of photos and generate a 3d Gaussian Splat scene in PLY format.

For example, this is a Gaussian Splat scene rendered in PlayCanvas.

What is a Splat?​

Gaussian Splat Scenes are not made up of polygons and textures. Instead, they are made up of many (up to millions) of individual, unconnected blobs called splats. A splat is just a particle in space with size, orientation, color and opacity.

Below you can see a single brown splat selected. The splat bounding box shows its orientation and size:

The gaussian part of the name comes from the shape of splat itself: the splat opacity has a gaussian falloff from its center to its edge.

Engine Support​

The PlayCanvas team has been adding support to the engine for loading and rendering Gaussian Splat PLY files:

Since the resulting files are often messy and require cleaning, we released SuperSplat, a tool for cleaning and processing gaussian splat PLY files:

PLY Format​

However, the default gaussian splat PLY format as exported by training tools is large.

This is because the uncompressed format stores a large amount of data per splat:

For example, the original guitar.ply scene file takes 132.8 MB (32 MB excluding spherical harmonic data).

Compressed PLY Format​

So we introduced a compressed PLY format for use in runtime applications. The compressed PLY file format ignores the unused spherical harmonic data and stores the rest of the elements in quantized integers.

The format can be summarized as follows:

• Split the scene into chunks of 256 splats
• For each chunk, store the min and max (x, y, z) for position and scale in floating point
• For each splat in the chunk, store a normalized and quantized value for position and scale (relative to chunk extents) and orientation and color

This data layout results in the following data per chunk:

And the following data per splat:

As a result, the compressed version of guitar.ply takes only 8.7 MB.

Do It Yourself​

The easiest way to generate a compressed PLY file yourself is using the SuperSplat tool. Load the PLY file into SuperSplat and export it again using the 'Compressed Ply File' option:

If you are interested in the file format specifics, see this code which demonstrates how to decompress the file data.

See this editor project for an example of loading and rendering a compressed gaussian splat PLY file. Or you can run it here.

Summary and Future​

We have introduced a new compressed PLY format for gaussian splatting which is roughly 4x smaller than uncompressed data and can be used in realtime applications.

In future we hope to:

• store splats hierarchically for optimized rendering and culling
• implement realtime splat LOD
• test skinning and animation of gaussian splats
• further compress gaussian splat data
• optimize WebGPU rendering

References​

The compressed format is largely based on the fine work of Aras Pranckevičius and his blog posts.

Today we're excited to announce that PlayCanvas now supports WebP and AVIF images natively in the editor. Two new, efficient image formats built for the web.

This is not just an exciting update for us but a game-changer for anyone looking to optimize their projects for faster load times and better quality.

JPGs and PNGs have served us well. They're reliable and you can use them everywhere. AVIF and WebP are the new kids on the block. They offer better compression with a wider range of features like alpha channels, HDR and wide color gamut. But how do these formats work in practice and are they really that good? Let's dive in.

Take a look at the following images:

They look strikingly similar, but the file size tells a different story. The AVIF image on the left has a file size of just 276Kb, almost half the size of its JPG counterpart at 508Kb! This level of compression is not just impressive on its own. When you're working on a project with multiple textures, switching to AVIF and WebP formats could considerably reduce your overall load time.

What does this mean for you? Well, that's simple: it's all about efficiency and speed without sacrificing quality. Textures are one of the biggest assets in a 3D project and AVIF and WebP significantly reduce image sizes compared to traditional JPEGs and PNGs and that means faster load times and happier users. 🙌

AVIF all the things?​

Ok, so browser support for AVIF is good, but not great according to caniuse, so always check support and use the right format for your users. Different formats are better suited for different content. It's not a one-size-fits-all solution, so we encourage you to explore various options and choose the one best for you.

Getting Started​

Easy! Just start uploading your AVIF and WebP files into the Asset Panel, and for those of you wanting to convert your existing assets, you can now right-click on your asset in the Asset Panel and quickly convert your asset into a PNG, JPG, WebP or an AVIF.

So there you have it! This is not just a feature release; it's our way of saying we're listening and we care. We're excited to hear your thoughts on these new features. Give them a try and share your experiences with us.

Till then, keep creating, keep innovating!

We're thrilled to announce the launch of the open source glTF Viewer 4.0, an update that supercharges your 3D model viewing experience with powerful features and support for the latest web technologies!

"Cyber Samurai" by KhoaMinh is licensed under CC BY 4.0.

TRY IT NOW

This new release is chock-full with enhancements aimed at providing more realistic, insightful, and versatile viewing options for your glTF files. Let's dive into the headline features of glTF Viewer 4.0.

New WebGPU Renderer​

Topping the list of today's updates is support for WebGPU! WebGPU heralds a new era in graphics and compute capabilities, offering enhanced performance and efficiency. Users can now select WebGPU as their default renderer, and don't worry if your platform doesn't support it yet - the viewer gracefully falls back to WebGL 2, and subsequently WebGL 1, depending on API availability. Note that WebGPU support is considered beta for the moment and you'll need to proactively enable it and refresh the viewer to check it out:

Also make sure you're running the viewer in a browser that supports WebGPU. At time of writing, this means Google Chrome!

Enhanced WebXR AR Mode​

Take your 3D models into the real world with our revamped WebXR Augmented Reality (AR) mode! Available currently on Android devices, this enhanced AR mode lets you view any model in your actual environment, complete with intuitive new controls that allow you to accurately position and rotate objects in the real world. Let's hope Apple decides to roll out WebXR support on iOS soon! 🙏

Frame Selected Node​

Navigating large scenes can be a pain - Viewer 4.0 addresses this by allowing you to select a node in the scene via the hierarchy panel on the left. You can then press 'F' on the keyboard to frame that node and recenter the orbit camera on that node's position.

Better Immersion with Projective Sky Dome​

"130" by mononofu is licensed under CC BY 4.0.

Experience realistic photographic skies with our new projective sky dome! While previous versions allowed for skyboxex with an infinite projection, 4.0 introduces a dome-shaped skybox projection that incorporates a flat ground plane. This warps the skybox texture to have a more believable appearance, delivering a more authentic and immersive perspective, melding your 3D models with strikingly realistic backdrops.

Debug and Inspect with Render Mode​

Ensuring that developers can seamlessly troubleshoot and inspect glTF files, the new render mode allows you to select and display individual inputs/outputs of the render pipeline, including albedo, emissive, normals, gloss, AO, and more. This new level of insight is invaluable for debugging, making it even easier to work with your glTF data.

Enhanced Realism with VSM Shadows​

The addition of Variance Shadow Mapping (VSM) casts your 3D scenes in a new light, literally! Shadows aren't merely aesthetic; they provide context and depth, especially in AR mode, assisting to ground your object naturally within its real-world environment. Explore scenes with a newfound depth and realism that draws viewers into the experience, both in standard and AR viewing modes. Find the new shadow controls in the Light Settings panel:

Join Our Open Source Community​

We're not just excited to share these innovations with you; we're eager to hear your thoughts and welcome your contributions! If there's a feature you're longing for, please don't hesitate to submit your requests.

Better yet, become an active contributor to our codebase! Our open-source community thrives on collaboration and fresh perspectives. So, dive right in, explore the code, and let's shape the future of 3D model viewing together! Your expertise and insights could help shape the next release.

GO TO GITHUB NOW

Conclusion​

With glTF Viewer 4.0, we're redefining the standards of 3D model viewing. From WebGPU-powered rendering to WebXR-powered AR, this update is designed to inspire, assist, and elevate your work with glTF data.

So stay creative, friends, and we'll see you on the forums! 👋

Today, we're excited to announce the integration of Sketchfab into the PlayCanvas Editor!

"Spartan Armour" by McCarthy3D is licensed under CC BY 4.0.

With today's launch you'll have instant access to Sketchfab's enormous library of high quality 3D content right inside the Editor.

What Is Sketchfab?​

Sketchfab is a platform that allows users to publish, share and discover 3D content on the web. You can think of it as a sort of "YouTube for 3D models". The platform hosts millions of 3D models in various formats, including glTF, OBJ, STL, and more.

Users can upload their 3D models to Sketchfab and embed them in other websites much like one would embed a YouTube video. This has made it popular for showcasing 3D models for a wide range of uses including game assets, 3D scans, architectural models, and educational content.

Accessing Sketchfab from the Asset Store​

Earlier this month, we announced the brand new PlayCanvas Asset Store. The Asset Store is an in-Editor panel of useful assets that you can use in your projects. With a few clicks, you can browse, discover and import various types of content. So it is the obvious place to incorporate the vast Sketchfab content library.

To access the Sketchfab library, simply hit the Asset Store button on the Asset Panel header and select the SKETCHFAB filter on the left.

Authorizing PlayCanvas to Access Sketchfab​

If you want to import models from Sketchfab into PlayCanvas, you first need to create a Sketchfab account. Then, the first time you try to import a Sketchfab model in the Asset Store, you will be asked to authorize PlayCanvas to access your Sketchfab account.

It's so quick and easy!

New License and Author Info for Assets​

Sketchfab assets are all tagged with a license and an author. When you import Sketchfab content into your PlayCanvas projects, it's important that we preserve this information. Now, when you click on any imported asset, you'll see we've added links to license and author information in the Inspector panel.

This ensures that the author's licensing wishes are respected and that they recieve credit for their work.

Go Forth and Create​

With today's release, it's never been easier to create interactive 3D experiences for the web. We hope you love the new Sketchfab integration. But, as always, we love to hear your opinions so head over to the forum and join the conversation. Happy creating! 🚧👷

Welcome to the new PlayCanvas blog! 😎

We have made the decision to move from WordPress to Jekyll and although it's early days, so far, we are very pleased with the results. Therefore, we thought it might be useful to explain the reasons why we did this and describe the process we followed.

Some History​

We originally started our blog back in February 2012 with a very low-key post. We installed our own instance of WordPress and that functioned fine for a few years. However, in 2015, we decided that we didn't want to deal with managing a WordPress instance ourselves and so we signed up for WPEngine, a WordPress hosting service. And there we stayed until today.

A Better Way?​

Over the years, some frustrations over WordPress did start to materialize. Customizing styling was tricky, bulk editing posts was not an option, managing media through the media library was quite inconvenient, and so on. On top of that, our blog was continuing to grow in popularity and WPEngine costs were rising. A team member suggested Jekyll as an alternative so we ran some tests to see what would be involved in a migration.

Migrating from WordPress to Jekyll​

First up, I should say that you should make use of ChatGPT when you attempt something like this. It made it a smooth and pleasant experience for me. But so you can see how straightforward it was, here are the key steps:

1. Export Your WordPress Content. On your WordPress Dashboard, navigate to Tools > Export and select 'All Content'. Click 'Download Export File'. You should receive an XML file that contains all of your site’s content.

2. Convert WordPress Export to Jekyll Format. To convert your WordPress export file to a Jekyll-friendly format, you can use a tool called exitwp. Here are the steps:

   • Clone or download exitwp from its GitHub repository.
   • Put your WordPress XML file in the wordpress-xml directory in the exitwp directory.
   • Run exitwp (requires Python and some libraries, see the repo's README for details).
   • This will generate a _posts directory containing all your blog posts converted into markdown format that Jekyll understands.

3. Setting Up Jekyll.

   • Install Ruby and Jekyll. For installation guide refer to Jekyll's official site.
   • Create a new Jekyll site by typing jekyll new my-awesome-site in the command line.
   • Replace the _posts directory in your new Jekyll site with the _posts directory that exitwp generated.
   • You can run your site locally to see how it looks by using the command bundle exec jekyll serve.
   • Then, open your browser to http://localhost:4000.

Once I had our blog post content successfully exported to a basic Jekyll site, I checked everything in to a new GitHub repository. From this point, I wanted any change I made to the content to be tracked via Git version control - I mean, why wouldn't you?

The initial migration to Markdown worked OK but it wasn't perfect. There was a lot of superfluous whitespace in the generated Markdown and syntax was broken in places. So I took the opportunity to do a wholesale spring clean of the Markdown content.

Prettier​

Prettier is a great code formatter but you may not know that it also has built-in support for Markdown formatting. I installed Prettier via NPM and ran it over all Markdown files:

prettier --write _posts

All formatting was now beautifully consistent.

Linkinator​

Linkinator is a tool for reporting broken hyperlinks in both Markdown and HTML. I scanned the locally generated Jekyll site with it using the following command:

linkinator . --server-root _site --recurse --verbosity error

I fixed well over 100 dead links. The older the post, the more dead links there tended to be (as you might expect). While I was always able to use Linkinator on the published WordPress site, it was never convenient to skip from post to post and edit hyperlinks in the WordPress UI (we have over 220 posts). So I never made the time for it. Now I was able to quickly search and replace links in Visual Studio Code and it was a breeze.

Aside from giving your readers a better experience, here's what ChatGPT has to say about dead links:

While Google's algorithms are complex and multifaceted, it is generally understood that having too many broken or dead links on your site can negatively affect your website's SEO ranking, albeit indirectly.

So this step was definitely worth doing!

Media Library​

Every WordPress blog has a Media Library that contains all of the images and videos referenced by your posts. The PlayCanvas Media Library had over 1GB of content. However, I noticed that a large amount of content was not actually referenced by any post. I wanted to ensure that I exported only the used media (since I didn't want to bloat the GitHub repository).

• Step 1: Install WordPress plugin Media Cleaner and use it to delete unused media files.
• Step 2: Install WordPress plugin Export Media Library to download all remaining media files in a ZIP.

This process cut the Media Library from 1GB to about 550MB. I simply extracted the ZIP contents to assets/media (relative to the root folder of my Jekyll site) and then did a search and replace of all media embed links to use the new file location. Easy!

Hosting on GitHub Pages​

I now had a fully exported and functional Jekyll blog. But where to host it? One option was to throw the static site onto a file server and put it behind a CDN. But wait, the file-set of the Jekyll blog now sits in GitHub, so why not use GitHub Pages to host the site? There are some great benefits:

• Deployment is a breeze to set up.
• It's a managed platform so you just don't need to worry about maintenance and security as you would with your own infra.
• And, oh yeah, it's free!

The Pros and Cons​

So now that we're migrated over, let's quickly summarize the key benefits:

• Free: Costs for our blog are now a pleasing $0.
• Secure: We have eliminated a potential attack vector. No more password reset confirmation emails!
• Powerful Tooling: There's a vast array of free and open source tools to lint and format Markdown.
• Easy Editing: Choose any editor you like such as the awesome Visual Studio Code.
• Versioned: All content is now under version control providing revision history, diffing and so on.
• Collaborative: We can take contributions to the blog via GitHub's pull request mechanism.
• Customizable: We have fine-grain control over the HTML and styling of posts using Jekyll themes.
• Platform Agnostic: Markdown makes it trivial to migrate to any other platform in the future.

OK, but what about the cons? 🤔 I have to tell you - I genuinely can't think of any! Sure, we don't have comments enabled now but as I understand it, Disqus has a Jekyll integration that we can turn on at some point. Seriously, feel free to ping me on Twitter if you think I'm missing something here. But I wish we had taken this step years ago.

Open Source and Open to Contributions​

As mentioned, the blog is now on GitHub. Take a moment to go check out the repository and hit that Star button if you like what we've done here. ⭐ We've taken the step to open source the blog's content under an Attribution-NonCommercial 4.0 International license. This is part of our continuing drive to open source most of what we do.

Ultimately, we want you to get involved and help us make the blog better and better over time. Consider submitting issues and pull requests yourself.

• Submit spelling and grammar fixes.
• Submit design tweaks and improvements.
• Submit your own posts showcasing your PlayCanvas projects.

It's your blog! 😊

That's all from me. Looking forward to many-a-future blogs posts on this new platform. 👋

The PlayCanvas Asset Store is the first place that users tend to go to find content for their projects. This is especially true for new users who want to get started as quickly as possible. Up until now, the Store has not been particularly easy to use and the content has not changed in quite a long time. In short, a complete overhaul and refresh has been long overdue. So today, we are incredibly excited to announce a major upgrade for the PlayCanvas Asset Store!

First up, check out a little example of building a city scene using content taken from the Asset Store. A skybox, a pack of 3D city block meshes and a camera control script are imported and the city is built via drag and drop. And not one single line of code is needed!

Now, let's examine some of the key highlights that make the new Asset Store so special.

Built Right In To The Editor​

It should be possible to grab assets quickly, right from within the Editor itself. Why should you have to open a new tab and go hunting around the web? So to keep things as convenient as possible, the ASSET STORE button (in the Editor's Assets Panel) now opens a nicely designed, responsive Asset Store panel.

One really cool benefit of selecting assets from within the Editor is that the currently selected Asset folder is known. This means you have compete control over where your imported assets will be saved.

Preview Store Assets Before Import​

Sometimes, an asset thumbnail just isn't sufficient to tell if a particular asset is what you want.

Our new Store allows you to select a Store Item and preview it in an appropriate viewer (glTF Viewer for 3D models, Texture Viewer for textures and cubemaps).

Find What You Want Quickly​

As the number of store items continues to grow, it's going to be incredibly important for you to be able to narrow down your assets searches. As a result, the new Store comes with powerful searching, sorting and filtering capabilities.

You can filter by asset type, search asset names and descriptions and then order search results on a host of criteria.

Fresh New Content​

Recently, the Asset Store content was, let's just say, beginning to show its age. We are now in a world of HDR skyboxes, PBR materials and high polygon meshes. So it made sense to erase old store items and refresh the Store with better, more modern content. We have selected a broad variety of Creative Commons assets from fantastic content sources such as kenney.nl, HDRi Haven, Sketchfab and Khronos' glTF Sample Models.

If you would have any suggestions for content you would like to be added to the Store, please do let us know!

The Future​

There's still so much we want to do with the new Asset Store! But here are some things we have in mind:

1. More Content. The important thing to say about today's update is that it delivers the core infrastructure on which we can iterate. It is now exceptionally easy for us to populate the store with more content. So, in the near term, you can expect to see the range of content expand quite rapidly.

2. Third Party Stores. Now that we have a solid foundation in place for the Store, we have the ability to host third party stores within the same UI and maximize your choice.

3. More Asset Types. Today's launch offers models, fonts, textures, skyboxes and scripts. Next, we want to add audio assets and template assets (AKA prefabs). Template assets in particular are very exciting because you would be able to import fully interactive, visual entities into your projects (such as a drivable vehicle or a controllable character).

What would you like to see us add to the Asset Store next? Let us know on the forum.

Happy creating, friends!

WebGPU is a cutting-edge technology that promises to revolutionize the way 3D graphics are handled on the web. As the successor to WebGL, WebGPU provides faster and more efficient rendering capabilities for complex 3D graphics and simulations.

PlayCanvas has been at the forefront of this new technology and has been working on adding WebGPU support to its platform.

With WebGPU, we can expect to see more immersive and interactive 3D experiences on the web in the future.

PlayCanvas WebGPU Clustered Area Lights Demo

Refactoring of WebGL engine​

Before adding support for WebGPU, it's important to discuss the significant amount of refactoring work that was required on our existing WebGL engine. Implementing deeper architectural changes while preserving backwards compatibility required a significant amount of meticulous care.

• To enable support for WebGPU, we needed to establish a clear separation of graphics technology that could be shared between WebGL and WebGPU. This involved a significant refactoring effort to extract WebGL-specific code into a separate set of classes.

• PlayCanvas utilizes a collection of shader chunks to produce GLSL shaders that implement advanced material properties and lighting modes, as well as custom shader chunks defined by users. However, since WebGPU employs the WGSL language, we used glslang and tint WASM modules to dynamically convert these shaders on-the-fly with injecting support for uniform buffers and other modifications.

• The PlayCanvas engine lacked explicit render passes, making the rendering process more rigid and harder to extend. This was solved by implementing a FrameGraph that allowed us to describe the rendering process as a set of render passes, their dependencies, and associated targets, which created a more flexible and performant rendering architecture.

• Unlike WebGL, which sets render state and shaders using a custom API, WebGPU specifies all those through render pipelines. To support both rendering APIs with optimal performance, we needed to refactor the render states into standalone objects that are efficient to compare and set up.

• To support the WebGPU platform, we need to undergo a significant refactoring to organize uniforms into uniform buffers.

• To facilitate the asynchronous creation of WebGPU device, we have introduced a new async API to create a graphics device, which is the primary breaking change required to adopt WebGPU.

What is left to do​

• Our primary objective is to align the WebGPU implementation with that of WebGL, and while we have made significant progress towards this goal, there are still some features that are missing. Furthermore, several smaller details require cleanup and rectification.

• We need to incorporate it into the Editor environment for both launched and published applications. Currently, only WebGL is available in this environment.

• Our primary objective is to achieve full parity with WebGL, but initially, we are focusing on basic implementations of some concepts to deliver a working implementation, which will need to be extended to meet our performance objectives.

• WebGPU provides developers with access to Compute Shaders, which enables more efficient parallel processing of data on the GPU. This feature can significantly improve the performance of complex algorithms and simulations, which may have been impractical to run on the CPU. With access to Compute Shaders, we can bring new visual features to the next level, such as advanced particle systems, post-processing and global illumination techniques.

Engine examples​

As an early pre-release of WebGPU, we have updated several engine examples to use it, which can be accessed on https://playcanvas.github.io. To use WebGPU, the Chrome Canary browser is required, with the 'chrome://flags/#enable-unsafe-webgpu' flag enabled.

PlayCanvas Examples Browser

Let us know what you think in the forums!

Attributions​

Oldsmobile Cutlass Supreme Sedan '71 by Barbo is licensed under Creative Commons Attribution