I found it was not very easy. You certainly miss the toolset from programs like Zbrush, Mudbox or 3D-coat. I think we need better tools for a first-person editor. We certainly will be working on that.
Meanwhile, here is the stone golem I created. It took me around 20 minutes to build.
Looks awesome! :) and it would be an really cool feature if you could make it walk around :)
ReplyDeleteBall brush makes ball men. That hand is p good tho.
ReplyDelete"Come to life"? Does it mean that we can expect a physics engine and a neuron-based control subsystem like in Framsticks, or something simpler?
ReplyDeletePhysics should be no problem.
DeleteThings you build can come to life. You can have canned skeletal animations. If you can match the skeleton (including size) to something out of the voxel data, that part of the data would bound with the skeleton and follow it.
Rigging procedurally in realtime is part of the challenge here.
Hmm... so more like in Spore? That would be amazing.
DeleteLooks cool :).
ReplyDeleteYou don't have any plans to release a tech demo. It doesn't even have to be interactive, just flying around in some random paths? Just want to see something moving in realtime.
Have been doing some progress on my own engine. I played around with some balls sculpting to and I think a realtime golem of that size could be possible. How many balls is he made of?
intresting, i would love such Golems, could they be made out of various stuff aswell? so maybe have wooden Golems, dirt Golems etc, this could certainly be intresting.
ReplyDeleteyour Project has much potential, and it could easily be a very big success, even when singleplayer only.
What i also wonder, would it be possible that, we say i build a house that i can save this houses blueprint and use it as asset? so that when i create future villages i can copy that house in? would make construction of large scale cities easier, maybe also make assets able to get imported, for example if a friend makes a cool statue or building he can send me a specific file and i can add it to my database.
Is it possible to get a higher voxel density on instances, or in this case, characters?
ReplyDeleteIt is possible.
DeleteNext question, moving voxel sets. Have you attempted this yet? Would it be easier/faster to lock in the generated polygons and link them to the skeleton as well? I would imagine having issues if the poly mesh is regenerated every frame. Not just for resource consumption, but if the mesh generates differently for every frame of animation you could end up with a stuttering mesh.
DeleteThis is really cool =P.
ReplyDeleteOne way you could do the Golem building, if you are using a set skeleton, is to display the actual skeleton, but then as something other than sharp lines, perhaps as sort of magic-y lines, to which you "bind" materials, This way you could potentially also make partially flying Golems and stuff (If you want), and it would allow people to make a proper Golem, rather than having to guess what the skeleton's like, and then failing miserably.
"I thought it would be nice if the creatures you sculpt could come to life. They could help you build or defend you against other creatures"
ReplyDeleteHoly shit, man. You dream big. Keep it up!
Pokestone? Golemon? =P.
DeleteInteresting Miguel, I am relatively new to actually trying to talk to my GPU but have been playing with voxels for a year or 2 now. I unlike you have pretty limited knowledge in the field and would love to see some sort of offering from voxelfarm that would give me access to technology like this.
ReplyDeleteI did get in touch by email but they basically told me they had no interest unless I was a big company.
Failing that I figured I already had some CPU based code but the real magic with voxels is happening on the GPU (this appears to be the bit you cracked), it looks like yet again you are at least 2 steps ahead of me ... http://blog.ccoder.co.uk/
Hi Paul. It is unfortunate we cannot share this project with more people at this time. We are planning on something to remedy that.
DeleteThis version of the engine uses GPU only for rendering. The procedural generation and voxel processing is happening on the CPU. It took us a long road to get there. We did experiment with different things. Using the GPU gave us some edge, but then we saw it was not easy to create code for that model. Maybe we could have managed internally, but as soon as the engine allowed others to create custom voxel layers, we would be passing all that complexity along. That I did not like, I wanted the engine to be fairly simple to extend. Also standard C++ code was more portable. If we wanted to have a farm of Raspberry Pi nodes making sure the GPU code would run there is not that simple. And last but not least, for most client applications the GPU will be already super busy rendering. There was no point in decreasing the quality of the graphics, or the framerate because of the generation.