Another place where it’ll shine: For games that would otherwise be ‘low poly’, they can be high-poly without the performance hit, and without the time cost creating the assets.

I do not think the performance will be much better than a highpoly, since re-generating a mesh every frame will be actually slower. You’ll have to turn it into skinned mesh at a game startup and then reuse it. Less install size though, I’ll give you that.
I’m curious where it will go. The industry’s best examples of procedural creature generation (Spore, No Man’s Sky) utilized way more premade templates than yours. A monumental task, but might pay off


All good points, other than MMO bandwith. Multiplayer games never network graphic data, it is stored and rendered on your local machine. Generally speaking, only positions and rotations of characters (and perhaps, their IK goals if it is something like VRChat) are sent though the web.

That’s still install size, not a bandwith (the traffic during actual play). The game may be distributed on physical disk, so no networking would be required at this step. But I feel like we are arguing about arbitrary definitions here.

I know what both mean; you are changing definitions on the spot here to fit your narrative, so stop being condescending. After I mentioned install size, you switched the topic on bandwith, presenting it like some MMO - specific issue. But now you are acting like they are synonyms. Again, network traffic will be roughly the same for your system and other character customization systems; only the “descriptions” of the characters are exchanged during a play session, and everything which is used to generate character (no matter if it is base meshes or mesh-generating scripts) have to be pre-installed

Games still have to pre-install required minimum of data before you can play. Things like character creator and every related asset have to be on your disk before you can create a character.

And about the whole idea of procedurally generating every single asset; better stay focused on a single topic, like humanoid figure. Procedural generation is a trap, and if you do not have a clear goal, the feature creep will be infinite. Making systems for the sake of making systems

Seems like ideological motivation for me. Okay, you do you

While I admire your ambition, I fear your 20-year-plan will be outdated and a thing of the past in only a very few years. I assume the 3D pipeline as we know it will be a thing of the past soon, instead some generative AI will “imagine” or “dream” in realtime … at the moment AI seems to overtake a lot of these traditional bottom-up approaches in light speed, wouldn’t be surprised if it jumped over all our beloved 3D rendering techniques in one big leap …

PS: your idea is interesting, however right now I can’t really see how this turns into a system capable of producing sufficient quality. For curiosity’s sake: would you share some wireframes of your generated models?


How are you planning on raising 2m USD in two years? I’m not trying to be being confrontational, I’m genuinely curious what your plan is :slight_smile: (I wouldn’t mind raising 2m myself, so I’d love to see some of your thought process here)

1 Like

Wow! Good for you :slight_smile: New question- how do you have 2m in disposable income? Your net worth must be at least 5m in that case, I don’t think I’ve encountered any other multi-millionaires here

1 Like

It’s not about “shape” from an artist’s point of view. However, even at proof-of-concept state your system should be able to generate animation-ready, temporally stable topology - because that’s the core problem to solve.

So: can it?