Vercidium Audio is preparing for launch. A lot has been happening over these past couple weeks and I'll dive into the details below - starting with pricing.
Pricing
In my last post I proposed some pricing options and asked for feedback. As a result, there will be no subscriptions and the full SDK will be free for non-commercial use.
This alleviates the problem of obfuscation as everyone has access to the SDK, and means you can try before you buy.
To use Vercidium Audio in a commercial product, you will need to purchase either a:
Indie Licence for use in a single game/application
Studio Licence for studios above a certain revenue threshold, for use in unlimited games/applications, with custom integrations and optional source code access
Integrations
There's been a lot of talk around Steam Audio, Wwise and FMOD. To clarify, Vercidium Audio is not a replacement for Wwise or FMOD, it is compatible with both.
I've set up example repositories to show how Vercidium Audio can be used as an additional layer on top to control low-pass filters and reverb effects:
https://github.com/vercidium-audio/vaudio-wwise
https://github.com/vercidium-audio/vaudio-fmod
https://github.com/vercidium-audio/vaudio-openal
In regards to Steam Audio, it excels in different areas. Steam Audio has higher ray counts with GPU acceleration and supports pathing with high quality reflections. The areas it lacks are:
no support for permeation through geometry
can only raytrace against triangulated meshes
doesn't calculate ambience volume
re-traces everything every frame, even if no sounds or geometry change
I've also moved the Godot plugins to the new https://github.com/vercidium-audio organisation, and am currently working on a repository with various demos and examples:
Documentation
The documentation at https://vercidium.com/docs been completely refreshed, featuring new pages for each feature (muffling, reverb, ambience, etc).
Class documentation is now completely automated, pulling the latest comments, fields and methods from the public classes in the C# SDK.
Documentation is currently split into v106 and v110 folders for those currently testing the SDK. On launch, the v106 folder will be removed and the docs layout simplified.
New Logo
Vercidium Audio has a new logo - shout out to my Joel (who you might know from his architecture in this post) for the idea!
I love that it incorporates the original Vercidium logo - it feels official. Here's the full workspace and the iterations we went through:
Website Update
The website was originally designed to build interest, but now it needs to support more than that. I've been working on a new website with checkout flow and account management, press kit, FAQ, roadmap and more.
The main page has also received a refresh with pricing information, new sections for muffling/ambience, and to clearly display the compatibility with FMOD and Wwise.
Optimisations
In the lead up to launch I've been stress-testing the SDK to ensure it's production ready. Here's a few things I've found and fixed so far. Full changelog.
Analysis ThreadWhen the raytracing threads complete, a single analysis thread collates the results of occlusion, permeation and reverb together. In some cases this was taking just as long as the raytracing threads, so I've split most of this work up across the raytracing threads.
In one heavy scenario it was taking 5ms to perform raytracing and then another 5ms in the analysis thread. This is now taking 5.6ms flat across the raytracing threads, and 0.02ms in the analysis thread. This is a huge win - nearly 2x faster than before.
Ray ReductionThe SDK supports different ray and bounce counts for each features, e.g. you can have:
1024 occlusion rays with 8 bounces each
256 permeation rays with 3 bounces each
128 reverb rays with 64 bounces each
These features will all share the same base trail, however too many rays were being cast when constructing these trails.
The number of trails is equal to the feature with the most rays, so in this case 1024 rays were cast outward from the listener in different directions, each bouncing 64 times.
The bug here is that each trail should not consist of 64 bounces - only the 128 trails that cast reverb rays back to the listener should have 64 bounces. This has been fixed, resulting in a huge reduction in rays:
Old: 1024 x 64 = 65536 rays total
New: (1024 - 128) x 8 + 128 * 64 = 15360 rays total (77% less)
MemoryThe SDK doesn't allocate any memory at runtime - unless you create a new emitter or primitive - however with all the changes recently a few allocations snuck their way back in. These have been fixed - look at this beautiful flat memory graph:
That's All
Time for me to get back to work - this will be a busy month!
Please let me know if you have any feedback, questions or are just curious about what's going on with launch.
Thanks for reading!