3D graphics compo - Discussion Thread
category: general [glöplog]
Hey all.
We want to improve the 3D compo, So we are throwing the topic here to discuses and collect more global opinions on it, outside of the sceners that setup the compo.
hopefully we can evolve it into a great compo that is both fun to enter for any skill lvl in 3D graphics and fun to watch.
So lets discuss. some questions that could help the conversation along.
- Have you been enjoying the 3D compo so far?
- What is stopping you from entering the 3D compo?
- What improvement would you suggest to make the 3D compo better?
We want to improve the 3D compo, So we are throwing the topic here to discuses and collect more global opinions on it, outside of the sceners that setup the compo.
hopefully we can evolve it into a great compo that is both fun to enter for any skill lvl in 3D graphics and fun to watch.
So lets discuss. some questions that could help the conversation along.
- Have you been enjoying the 3D compo so far?
- What is stopping you from entering the 3D compo?
- What improvement would you suggest to make the 3D compo better?
what is "the 3d compo"?
One thing I’d like is a few more options for material setup - very specifically, I’d like to be able to use an emissive shader / no lighting, so that I could make a completely handpainted scene without influence from realtime lights :)
Apart from that .. perhaps the 100k triangles rule is a little too restrictive. I’m not saying there should be no limit at all, but maybe 10x that limit?
Might be fun with possibility for animation as well? Both baked animation and camera motions?
Apart from that .. perhaps the 100k triangles rule is a little too restrictive. I’m not saying there should be no limit at all, but maybe 10x that limit?
Might be fun with possibility for animation as well? Both baked animation and camera motions?
So here are a few random thoughts as someone who's basically responsible for this: My initial cursepiration for this competition was the obvious - the scene is lagging behind on high quality 3D, and if there are indeed 3D artists in the scene who are great (and I know there are), they don't have the same level of showcasing ability as other graphics artists who have several competitions at their disposal. Sure, a 3D artist can enter a 2D compo with a rendered image, but it's not the same: you can postprocess, you can paint over, your meshes can be millions of triangles, you can use high-end physically correct rendering, and it just has to look good from that one angle, as opposed to "realtime" 3D where there are very real constraints. Sure, they can still contribute to demos, but I've seen many attempts for that fail simply because there's a lack of bridging either because the coders weren't interested (because they wanted to stick to fractal 4ks) or because there wasn't good enough tooling / engine features.
So the context here is that I thought it was worth trying a "bridge-compo" that allows artists who work with realtime 3D (game artists, etc) to have some visibility as well.
With that in mind, a couple of thoughts:
- Oni and I had a lot of discussions about what level of featureset should this compo constitute of - lighting / camerawork / animation being the most discussed one. My current opinion (and I'm not entrenched in this) is that the closer you move to that, the more it just becomes another demo compo, at which point it loses its meaning a bit. I think the fun part of the compo as it has been at e.g. Evoke is that the focus of an entry becomes the actual raw asset quality (modelling / posing / texture work) - lighting / animation may still fit into this, but I feel adding camerawork would turn the 3D compo back into a 2D compo, and at that point you might as well just make a demo? I don't know. My thoughts on this aren't fully formed, and they're largely driven by the holistic view on things outlined above, rather than practical considerations.
- I've been very outspoken about how I considered using Blender as a viewer a mistake at this Revision, and I stand by that: first, I think it makes the compo partial to that one editor, where people using Max or Maya are now on a back foot, because Blender Eevee is considerably overpowered (procedural geometry, higher rendering quality etc.) - as predicted, it did indeed become a Blender compo. And make no mistake, if that's what you want, fine, but I think you should rename it to "Blender compo" then. (Which honestly then becomes the whole "are we associating with / endorsing Blender" type of problem.)
- In terms of shader features, it's obviously a work in progress, cce did a heroic job making the thing work, and there's always room for expansion like the aforementioned emissive maps or opacity maps, but part of the reason for this thread is that we realized that there's been no discussion, stuff just sorta happened and while I think we've proven that the compo has its place (and I do wish more parties would adopt it), the thing is so malleable that giving a bit of thought can go a long way.
So the context here is that I thought it was worth trying a "bridge-compo" that allows artists who work with realtime 3D (game artists, etc) to have some visibility as well.
With that in mind, a couple of thoughts:
- Oni and I had a lot of discussions about what level of featureset should this compo constitute of - lighting / camerawork / animation being the most discussed one. My current opinion (and I'm not entrenched in this) is that the closer you move to that, the more it just becomes another demo compo, at which point it loses its meaning a bit. I think the fun part of the compo as it has been at e.g. Evoke is that the focus of an entry becomes the actual raw asset quality (modelling / posing / texture work) - lighting / animation may still fit into this, but I feel adding camerawork would turn the 3D compo back into a 2D compo, and at that point you might as well just make a demo? I don't know. My thoughts on this aren't fully formed, and they're largely driven by the holistic view on things outlined above, rather than practical considerations.
- I've been very outspoken about how I considered using Blender as a viewer a mistake at this Revision, and I stand by that: first, I think it makes the compo partial to that one editor, where people using Max or Maya are now on a back foot, because Blender Eevee is considerably overpowered (procedural geometry, higher rendering quality etc.) - as predicted, it did indeed become a Blender compo. And make no mistake, if that's what you want, fine, but I think you should rename it to "Blender compo" then. (Which honestly then becomes the whole "are we associating with / endorsing Blender" type of problem.)
- In terms of shader features, it's obviously a work in progress, cce did a heroic job making the thing work, and there's always room for expansion like the aforementioned emissive maps or opacity maps, but part of the reason for this thread is that we realized that there's been no discussion, stuff just sorta happened and while I think we've proven that the compo has its place (and I do wish more parties would adopt it), the thing is so malleable that giving a bit of thought can go a long way.
One more quick thing: I really hope the discussion is centered around the craft of things and the compo's potential place in the scene, rather than how the viewing experience is for an attendee and what they wanna see as a bystander.
the emissive shader request came out of my recent 3d entry at Evoke where I had a specific idea in my head of what I wanted to do (stylewise) but couldn't.
re: the camera setup and animation - I guess I see it as a way to gain a bit more control of how you showcase your work. The presentation of the thing, so that it isn't exclusively left in the hands of the graphics compo organizer. Like when you setup things if you want to showcase something on instagram.. pick the nice camera angles, etc. I want to stress that it was just a thought though - not something I'll fight for at all... (might also turn the development and maintenance of foxotron into a nightmare.. it's nice to keep the tool as simple as possible)
re: the camera setup and animation - I guess I see it as a way to gain a bit more control of how you showcase your work. The presentation of the thing, so that it isn't exclusively left in the hands of the graphics compo organizer. Like when you setup things if you want to showcase something on instagram.. pick the nice camera angles, etc. I want to stress that it was just a thought though - not something I'll fight for at all... (might also turn the development and maintenance of foxotron into a nightmare.. it's nice to keep the tool as simple as possible)
I was expecting something like what you can see on sketchfab, with the same features in lighting, effects, etc..
Thanks for reading the rest of the thread, rez.
Quote:
I was expecting something like what you can see on sketchfab, with the same features in lighting, effects, etc..
sketchfab already has those features, so if that's the goal, let's just use that instead of trying to recreate sketchfab :)
+1 for the Sketchfab route. I agree allowing too much tinkering with the viewer to allow more customization in materials/shaders/procedurals ("the Blender route" in Gargaj's post) will too much distract from the 'a nice 3D model'-craft. Before you know it someone will model a cube and write some material with a nifty vertex shader and what not to morph it into a dragon and that's not "3D modelling" ;) So, simple PBR materials (including emission) + simple camera (fixed fov, no dof, etc) +3-point (individually colored) light setup with option to toggle shadows for each should be enough for everyone. No further fancy lighting from the viewer, if you want AO then just bake it.
As for what's stopping me... I was working on smth way too overambitious for Revision with several sculpted humanoids and scattering glass, will try to finish it for next year then (or more likely the year after) :P
As for what's stopping me... I was working on smth way too overambitious for Revision with several sculpted humanoids and scattering glass, will try to finish it for next year then (or more likely the year after) :P
so, based on my personal tastes, since demos and demoparties compos are mostly based on limitation I would be more interested in a 3d competition with stricter rules like low poly mesh with really small textures (like a single 256*256 picture).
with these limitations it would be easier to judge the entries and I would be more interested in participate because I feel it's more doable in a decent time than a "no limit" competition.
with these limitations it would be easier to judge the entries and I would be more interested in participate because I feel it's more doable in a decent time than a "no limit" competition.
i disagree with that as it's not 1996 anymore, so a ~150k poly mesh with ~100MB worth of textures (variably dimensional as the artist sees fit) is not uncommon for a central piece 3D asset in a modern game. if you like rez's suggested limitations, go bloody model smth for an Amiga demo instead :P
I think there's space for both, the same way pixel compos can coexist with high end graphics, or C64 compos can coexist with PC. I personally do like the idea of a very low polycount compo and there are probably parties where it could fit.
yeah and someone can still submit a 100 poly floppy disk with only a 256x256 texture in said compo if that's their vibe. just shouldn't be the top bar for a regular/general 3D compo though
Oh sure, but you can easily reflect that by just calling it "Low Poly 3D compo" or something like that.
agree with having both - I'd love to do a lowpoly 3d entry as well as a more wild almost-no-holds-barred 3d scene entry
So for me the 100k triangles was fine, but I would not mind doubling that budget. My problem at Evoke was the 50mb zip file size. I had a lot of textures and space was running out fast. So setting that to 100mb would be great. I also hope we get a proper opacity map function so I don't need png's with transparency xD
I am also very in for a low poly 3D compo. I think that would be a lot of fun, but I do have a feeling that we might first have to grow participants in the 3D compo before we could fill both a High end 3D compo and a Low ress 3D compo with enough entries to make it a compo. But if a party would make a Low ress 3D compo ill join for sure :)
As for animation/camera tracks after all our talks and also reading through the conversation here I do think that can spiral out of control fast. When I was thinking about animation I was thinking of things like VFX. but if we would allow a type of animation, it would just turn into discussions why certain type of animation are allowed over others. So my 5 cents on that would be lets not allow any.
But I do think presentation is important. I don't mind if my topology is not perfect, I just want to make something that looks cool. and so far I have always had to decimate my models anyways to fit within the limit. So any type of nice topology I had went out the window during that step of making it compo ready.
So I do hope that we get some controls on light sources. I think a 3 point light setup would be great like Pablo mentioned. that you can give your work a theme or a nice rim light of your choice at least. Or maybe a customizable Lightdome?. Few gradiant sliders where you can swap colors for the sky and ground plane and a sun/shadow direction you can control. If we stick to Foxotron it would be great if that could get build in.
I do think right now Foxotron is the most compo organizer friendly tool atm. There won't always be graphics entry orgas that are familiar with 3D software. So for them to just drop and drag a file and it works is great. It would be nice if we could have a save function that spits out a config file. So that we can just upload that together with our entry and everything looks the way we want it to in the viewer. I haven't looked into Sketchfab, but don't you need internet access for that. What if a party doesn't have internet ?
The whole blender topic is kind of closed on my side. After talking with you Gargaj and other people I can see it is giving a more favorable hand to people using blender and setting it up to work as a viewer without menu's is also not very Orga friendly unless you know your way around blender.
My main thing is how can we can get more people interested and more people participating. During the revision 2022 unofficial blender compo we definitely had some very beginner people that joined in the compo, most likely because it was blender. And I do think that was nice and hope those people keep participating and growing in their craft even when we don't allow blender files
During evoke we were talking about maybe building either a website or expending the Foxotron wiki with maybe some video tutorials on how to setup and export your files properly and even going into more depth about how to bake a high ress model to a low ress mesh. Do you peeps feel that would help get some more people involved and entering the compo ?
I am also very in for a low poly 3D compo. I think that would be a lot of fun, but I do have a feeling that we might first have to grow participants in the 3D compo before we could fill both a High end 3D compo and a Low ress 3D compo with enough entries to make it a compo. But if a party would make a Low ress 3D compo ill join for sure :)
As for animation/camera tracks after all our talks and also reading through the conversation here I do think that can spiral out of control fast. When I was thinking about animation I was thinking of things like VFX. but if we would allow a type of animation, it would just turn into discussions why certain type of animation are allowed over others. So my 5 cents on that would be lets not allow any.
But I do think presentation is important. I don't mind if my topology is not perfect, I just want to make something that looks cool. and so far I have always had to decimate my models anyways to fit within the limit. So any type of nice topology I had went out the window during that step of making it compo ready.
So I do hope that we get some controls on light sources. I think a 3 point light setup would be great like Pablo mentioned. that you can give your work a theme or a nice rim light of your choice at least. Or maybe a customizable Lightdome?. Few gradiant sliders where you can swap colors for the sky and ground plane and a sun/shadow direction you can control. If we stick to Foxotron it would be great if that could get build in.
I do think right now Foxotron is the most compo organizer friendly tool atm. There won't always be graphics entry orgas that are familiar with 3D software. So for them to just drop and drag a file and it works is great. It would be nice if we could have a save function that spits out a config file. So that we can just upload that together with our entry and everything looks the way we want it to in the viewer. I haven't looked into Sketchfab, but don't you need internet access for that. What if a party doesn't have internet ?
The whole blender topic is kind of closed on my side. After talking with you Gargaj and other people I can see it is giving a more favorable hand to people using blender and setting it up to work as a viewer without menu's is also not very Orga friendly unless you know your way around blender.
My main thing is how can we can get more people interested and more people participating. During the revision 2022 unofficial blender compo we definitely had some very beginner people that joined in the compo, most likely because it was blender. And I do think that was nice and hope those people keep participating and growing in their craft even when we don't allow blender files
During evoke we were talking about maybe building either a website or expending the Foxotron wiki with maybe some video tutorials on how to setup and export your files properly and even going into more depth about how to bake a high ress model to a low ress mesh. Do you peeps feel that would help get some more people involved and entering the compo ?
In terms of lights, would you (plural) prefer loading lights from the mesh file - with the implied assumption that there probably will be differences between the lighting model of your modelling tool and the shader we use - or would you prefer if Foxotron had a few stock lighting setups?
we talked about this briefly at evoke https://youtu.be/r-VGfN-4XKo
my take on the subject is that we should be aiming at 3d communities outside the demoscene and inviting them to participate as a gateway to meet demomakers and eventually collaborate with some.
to answer the questions directly:
- Have you been enjoying the 3D compo so far?
it feels like it either wasn't very well advertised or people been shy from participating in it, low number of entries overall. it's nice idea but doesn't seem like it's reached it's proper potential yet.
- What is stopping you from entering the 3D compo?
i suck at 3D modelling
- What improvement would you suggest to make the 3D compo better?
bridge out to 3d modelling communities, via their forums etc with an honest invitation to have them participate to show the demoscene community their skills and eventually partner up with demomakers.
my take on the subject is that we should be aiming at 3d communities outside the demoscene and inviting them to participate as a gateway to meet demomakers and eventually collaborate with some.
to answer the questions directly:
- Have you been enjoying the 3D compo so far?
it feels like it either wasn't very well advertised or people been shy from participating in it, low number of entries overall. it's nice idea but doesn't seem like it's reached it's proper potential yet.
- What is stopping you from entering the 3D compo?
i suck at 3D modelling
- What improvement would you suggest to make the 3D compo better?
bridge out to 3d modelling communities, via their forums etc with an honest invitation to have them participate to show the demoscene community their skills and eventually partner up with demomakers.
@ps this is not the opinion compo thread, so maybe put a pin in that until your area of expertise is in demand? :)
@darya erm, no thanks?! if you don't like reading my opinions you're welcome to ignore them, there are userscripts for that. i'll keep giving them whenever i deem them relevant to the topic at hand.
Quote:
In terms of lights, would you (plural) prefer loading lights from the mesh file - with the implied assumption that there probably will be differences between the lighting model of your modelling tool and the shader we use - or would you prefer if Foxotron had a few stock lighting setups?
I think the most realistic thing is to have some stock lighting setups to choose from within Foxotron (because of what you mention.. if the lighting model wouldn’t match exactly anyway, I imagine the process would be a lot of back-and-forward to try and get Foxotron to match whatever you had in mind)
But again… emissive shading? Then you could just bake your lighting into textures and not worry about it :)
yeah but then you're wasting pixel data thus file size for something that can be done cheaply real time plus it's just common practice to have your lighting real time in games. i'd say stick to what the industry does, the easier it also is for non-demosceners to contribute.
so as for lights: i think what Gargaj said you indeed need to set this up in the viewer. Can be rather simple: just some dialog to add max 3 lights (point lights? but that's just my personal preference as i hate directional lights :)) perhaps even some gizmos around them to custom place them in the right spots (it's usually sun, main light, backlight).
What sketchfab does with 'points of interest' for the camera may be what Oni wants. then you can pick some sexy artist-predetermined angles to show the scene. can be just multiple cameras defined in the fbx, can be multiple cameras you setup in foxotron if you're doing the light setup already there as well, even steven. Obviously the compo organiser should still orbit around the scene and also toggle wireframe mode to show how it's all done for good measure.
and we should not allow animation.
so yeah, use foxotron.. i don't think we should use sketchfab as-is, it's just, they already invented the wheel for a web-based 3D viewer that has enough features to showcase game assets.
so as for lights: i think what Gargaj said you indeed need to set this up in the viewer. Can be rather simple: just some dialog to add max 3 lights (point lights? but that's just my personal preference as i hate directional lights :)) perhaps even some gizmos around them to custom place them in the right spots (it's usually sun, main light, backlight).
What sketchfab does with 'points of interest' for the camera may be what Oni wants. then you can pick some sexy artist-predetermined angles to show the scene. can be just multiple cameras defined in the fbx, can be multiple cameras you setup in foxotron if you're doing the light setup already there as well, even steven. Obviously the compo organiser should still orbit around the scene and also toggle wireframe mode to show how it's all done for good measure.
and we should not allow animation.
so yeah, use foxotron.. i don't think we should use sketchfab as-is, it's just, they already invented the wheel for a web-based 3D viewer that has enough features to showcase game assets.
Quote:
yeah but then you're wasting pixel data thus file size for something that can be done cheaply real time plus it's just common practice to have your lighting real time in games. i'd say stick to what the industry does, the easier it also is for non-demosceners to contribute.
you're missing the point.
if I want to make that stylistic choice, it would be cool if I could make it - not be subjected to some arbitrary judgement that realtime lighting is better because games, right? :)
well, if we aim for PBR as e.g. implemented in Blender, Max or Unity then it will have emission in its material/shader which simply put is just a lighting override and you can be as stylistic as you want... hell, you can even omit lights from your scene and light your model with baked emission maps if you want. it just shouldn't be the default light implementation as most other artists most likely prefer to have their models real-time lit :)