MakeHuman for IMRSV Tools (VR Compositions)

[IMRSV_Pete]

Hello community.

We are working on a VR composition and distribution platform for short-format immersive experiences.
It is built on top of Unreal Engine 5.5+, but uses the open formats of Universal Scene Description (USD) and MaterialX for all asset content to ensure the effort put into them has a life before and after our project.

I'm thrilled to see continued life in the MakeHuman project especially in recent weeks with @Joel posting more tutorials.
This project is the best hope for an open source character creation tool were artists are free to do with their creations whatever they see fit without restrictive distribution language forcing compliance to closed ecosystems and formats.
Refreshing in a world were orgs are trying to lock creators into dead-end marketplaces and bespoke implementations with subpar exchange formats.

Our project will be dependent upon the creators' ability to freely share their projects in order to contribute to a robust ecosystem of assets from unique skeletal meshes to wardrobe and makeup, accessories and more.

We see this with non-character assets as well and already use Blender (with USD export) for "Set" and "Prop" creation.
I can say with confidence - everything looks great in our project (Unreal Engine with Lumen and Megalights) even on first import.
I have tested the Blender USD export flow with a MakeHuman Skeletal mesh and this does work.
With some collaborative thinking of peeps who understand the dynamics a bit more - we could streamline this experience and make it an interesting new path for some users of MakeHuman and Blender.

We are not experts of character creation, skeletal meshes, skins, hair, clothes, etc. and can really use community insight and guidance to get to the right high-quality, realistic result we seek.

I would love to connect with any devs in this community who would consider commission work to help further the MakeHuman project as well as help us streamline the pipeline from MakeHuman/Blender into our composition tool.
Meaning - the contributions made to enable our project are also contributions to this project.

Would this be the right place to have that discussion?

[RobBaer]

I am not such a person, but this would be a perfect place to find such individuals. Welcome!

[tomcat]

This project is the best hope for an open source character creation tool

The idea is good, but the choice of engine is strange. Maybe it is better to use an open source engine for such a project?

[IMRSV_Pete]

the choice of engine is strange. Maybe it is better to use an open source engine for such a project?

Thanks for the feedback. I can see how the choice for Unreal can feel odd if the thinking is only about open solutions.
But we also need it to look and perform great.

I should also clarify - we are not starting the project - we are close to an Alpha release of IMRSV and seeking the most appropriate asset pipelines (including characters) to keep these as open and community driven as possible.
(That is where the cool future stuff happens)
The engine discussion ship sailed for us 2 years ago.

That said.
We did look at everything we could find at the time (including open source) and considered building from scratch on top of a few promising open source pure render engine libraries.
(In hindsight - that would have been a mess)
We actually started the project on another game engine and had to switch because fidelity and performance was just not where we needed it to be.

We want normal people to be able to create realistic experiences quickly and easily with little to no 3D knowledge. (Shaders, baking, cooking, UVs, etc…are a no go).
That means you add a light to a scene and it works like a light without baking, probes, or hidden bounce lights, etc.
I have not seen anything that looks better and is more performant with raytracing in VR than Lumen for realistic scenes.

The things you can do with a high poly architecture USD using an HDRI and a single light is stunning in VR while still running 120fps with a gaming PC.
Now with megalights - we can add another 100 lights and it's still just as performant.
We want to extend that experience to characters knowing this is WAY more complex then architecture.

Bespoke assets customized just for our project - also a no go.
It has taken quite some effort on our side to sit assets outside of the Unreal Engine “cooked asset” approach which is why we chose USD with MaterialX.
These assets don’t get locked away in to PAKs, CookedAssets with bespoke logic attached, etc
We pull them in at runtime.

To achieve that magic - we commissioned the development of a “USD Runtime” Unreal Plugin with that community and will be opening it up very soon for others to do with as they please.
That is an example of a project we helped support with commission work similar to what I propose here.
That project (and a few other connected open source projects) benefited and as a community interested in open pipelines win.

It’s not about building for our specific project that uses Unreal as a VR render engine.
It is about building the right long term solution for a specific use case.
In this group - it is about realistic characters (and related assets) that can be freely shared and evolved.
Inspiration to take a project forward into what feels like “the right way” can be hard to justify for open source projects that are trying to stay connected to legacy needs.
This is why we can help with an actual pipeline / use case and some financial support to think what could be next.
It’s not theoretical. We have a practical use case that can be be tested right now if people are interested.

I really want to support the establishment of open pipelines - even if the tools they go into/out of may not be - yet.
As Godot, O3D, and others explore USD and MaterialX, make incremental progress on realtime visual fidelity, and some progress on VR support (which is just not a prio for those communities right now - I asked) ->We want to get this powerful tech in the hands of immersive creators now so we can learn what is REALLY needed to take immersive technology further in the entertainment space.

So that is why we use Unreal as our realtime render engine and USD/mtlx as our asset transport and composition format.

If you have suggestions that perform at that level for VR today - we are certainly open to exploring it for the future.

Again - our project is USD and the compositions (project files) are USD.
They would run on ANY engine that can read USD.
If it looks as good or better than Unreal - we will not ignore it.

Long answer - but hopefully one that helps spark some understanding and interest within the group.

[tomcat]

I can see how the choice for Unreal can feel odd if the thinking is only about open solutions.

I understand your position. Recent events with Unity have demonstrated the great danger of relying on proprietary engines.

But we also need it to look and perform great.

“To look” depends more on the assets than on the engine.

However, it's your choice and your risks.