Official Whiteboxing

I think we might rather allow viewing far away stuff only from specific locations - you can't see it, it's no problem it isn't correct. I would also prefer a 2D Image at the correct location (with transparent background for smooth transition to the skybox) because that way it doesn't follow you around like a skybox. This way it is further away when you are further away and you view it higher or lower when you are at different height. You can see this a bit in my 2 scene load demo (

) albeit those squares are just 20 m wide.

Actually when playing so far I enjoyed the jumping up those "stairs" and with using 2D images for the far away view we get away with more. Also high walls have the advantage of blocking the view. You can't see the town from the beach and I think this is a good thing. So you focus on the tutorial. But as at the end of the tutorial the path to the next part is opened in my opinion you would naturally go where you couldn't go before. No need to see the town already. Soon after I think there should be the view - telling you are on the right path (beside being as beautiful as we can possibly make it). I like @itsLevi0sa 's solution with walking through that stone arch. It was on the list of things I was looking forward to model.

The original reason for having a view at the town from higher up was that the default camera angle is approximately 26° out of the horizontal from above. Of cause you can rotate the camera, but as long as you don't know there is something to see, why should you? Being not entirely satisfied with view of the town in my first attempt of roughly wideboxing the reduced map, I spend my morning on what would be called a full factorial DOE if the results were numbers. In this case the results are screenshot's of the view that you get with a camera at a defined position (the camera set to field of view 40 the same as the game camera):
As you see in the default view column, that at 250 m distance to the towncenter 15 m above the towncenter is kind of the absolute minimum height from which you can see something of the town in the default view (at 150 m it is a little less). I think 40 m height above the towncenter is not really an option, but here you see that the town would be almost entirely visible from the default view should we want that.

I like the far away view at 250 m better. Interesting is too that rotating around the towncenter in the horizontal makes a difference and can help us as well. This is because the town isn't circular but extends more in the direction of the road from which we reach it.

Besides I think that second mountain, which was in my previous whiteboxing but is not in map right now, does a lot for the view. It is one of those things that we might do as inaccessible terrain.

 

That's an original idea indeed. But I think it might be problematic in the long run. For instance, what happens if you're close to the bridge and you jump?

And also, we need to find a solution which kinda works for every situation. I think the bridge part is relatively easy to solve, but I'm quite worried about the town, which will be surrounded by terrain on all sides (i.e. requiring a lot of geometry to be 3D rather than skybox). But we can figure it out.

So you'd also be ok with a small reduction in size of the island?

 

Thanks for the detailed analysis, Smurjo.

This can work up to a certain point, for sure. But we will need to build a pipeline to produce these images which, as we change the map, will need to be reshot again and again.
Having a Prefab with 3D geometry blocking most of the view + a skybox for the background is a bit more maintainable.

I have no issue with what's in the whiteboxing, I also find the progression quite interesting. I'm mostly talking about scale (in terms of gameplay) and very connected to it, the amount of details we can push if the scale is too high.

 

So... are you changing the whiteboxed map in the end? I would like to have a map we consider "final", at least in terms of size and position of locations. We can then tweak roads and playable areas, but this would give us a rough idea of how big the game world needs to be.

Also small thing: I've merged

main

to

whiteboxing

, so you have all the tools and art in

main

. Pull the latest

whiteboxing

to get it.

 

Your mean this map?

I wouldn't have proposed it if I couldn't go with it.

 

Yeah, I wasn't sure if you yourself were convinced about it. I always like to confirm. But if you agree with me and @itsLevi0sa seems to too, let's run with it.

I was also going to use the map to show something in tomorrow's livestream. Given the short timeframe, I'll probably use the current one, dress it, make a test, and then we'll redo it on the revised smaller map once it's ready.

 

Of course, whatever works best! I'd go too with @Smurjo 's improvement. That was some good documentation you did there by the way, I like the detailed approach too!

 

I have done a video on my whiteboxing of the reduced map so far:

It is textured with the map.

 

I think the scaling is much better, perhaps after putting more assets we could add/reduce some areas a tiny bit still. I have the feeling that everything is too close now, but perhaps I need some time off to get "detached" from my own approach. But if we go with scenes loading (fade in and outs) perhaps we could detach those areas a bit and add some fillers (those prefab walls) just to make them feel bigger even though the gameplay walkable area is exactly what you've shown. Did you implement your scene loading solution on this video?

 

I feel the same about it too, since I played it several times I was getting used to the space. But of course it takes a lot of art assets to fill.

Not yet, it is still one scene. I wanted to have some confidence in it before I carve it up.

I did this with the old map though. It worked without problems. There is some change of light when you load or unload a scene with a directional light in it. So you either have to put the directional in the permanent scene or put some effort into the smooth transition of the light at the scene border.

 

What is the beginning? Is it supposed to be the beach?

I'd like to see the map. There are a couple of things that worry me, for instance that mountain path is immediately above the town, which means we'll have to have the whole town as a 3D backdrop? That might be a lot to handle (it's basically having 2 scenes loaded at the same time).
That's why I was originally suggesting the idea of teleporting between areas that are more distant, to have Location areas and "not much" in between, which would help with clearly separating them from each other and with creating lighter Prefab walls (they don't need to obstruct the view, they just need to provide the feeling that there are 3D objects in between).

 

I still think that the combination of what you both suggested previously would be ideal. A big map but with small playable areas that you teleport in between. Otherwise a seamless walk would mean we would shrink everything at the cost of losing some nice distant overviews. So the mountain path could be something completely cutoff from the town, you would see it in the distance, you would know that a particular town exit would lead there, but you'll never have to walk it all. Then the mountain perhaps might have no overviews of the rest of the island at all, perhaps only at the end so you know where the exit path will lead you, but again, not really necessary. I would like to experiment with those skyboxes more in the different scenes, I didn't know about the use of reflection probes as skyboxes and I find it brilliant.

 

I have just raised a PR for the reduced size map: https://github.com/UnityTechnologies/open-project-1/pull/330

I have also reworked the visibility of the town from the mountain path, now featuring

• an additional waterfall
• an outlook to the ocean
• and a dedicated spot from behind the lake, where the town can be seen (whereas it can't be seen in-between)

Whatever the approach - the far away views would have to be created. I think we could even cheat and have a far away view which seems further away if we loose vision of it before we actually reach it.

 

Tonight I've shown some set dressing stuff and brought some considerations to the livestream:

1:23:26

Soon I'll push it all to

main

and other branches.

In other "news": since

main

and

whitebooxing

are pretty much aligned, can I remove the second? You'd be making PRs to main directly, so you're always using the latest stuff.

 

@cirocontinisio:
I watched the livestream yesterday - in fact I look forward to the livestreams every time. I consider them like the next best thing to regular project meetings (I would like to say they are more entertaining than project meetings, but that isn't true - I have had many very entertaining project meetings as well).

It was a little confusing to me though - those scenes were from the old map (easily recognizeable by the scale). And it made clear to me that the scenes are too big for the number of different art assets we have - even turned and scaled the decorations become repetitive. They are also too big for the number of "attractions" we have - there aren't that many NPCs or things to find and you can fight the same sort of critters only so often before it gets repetitive again. I think we need a decision on the scale, I'd really like to finish the forest.

It also didn't escape my keen eye that you had mesh colliders on all the rocks even in the back drop scene - all of them with the full mesh of e.g. for the big rock 748 triangles (the complete ground of my forest whiteboxing only has 692 triangles so far, and I was thinking of keeping the whiteboxing mesh for the collider, even if I intend to round out the edges and put more details on the render mesh). Albeit you emphasized the backdrop scene was much simpler, it really was close to a second scene - and been in potentially several different scenes it will be quite some effort to keep the backdrop stuff aligned with the actual scene it is showing.

Which leads me to proposing a 3 scenes approach: 2 backdrop scenes at a time and a "details" scene in the center of the scene where the player is. When teleporting (I am still not fully convinced how to sell that in a logical way) you only load the backdrop scene where the player lands at first. Then you load the backdrop he can see (but he lands facing into the scene we already loaded... hoping by the time he turns the other scene is loaded) and then the details (With the teleport target a little off the detailed part of the scene).

It's a cool method to create the skybox by reflection probe. I think we can automate it even more using the near plane property of the reflection probe instead of taking out the near scenes. If we go with the 3 scene approach we would only need to load all the backdrop scenes for baking the skybox probes, not the detail scenes.

Very good news for me, I have had trouble with making a PR to whiteboxing yesterday (800+ commits behind?), so I made my PR for that small scale map to main.

 

Yes, in fact I think we will throw away completely what I've shown in the livestream. That's why the day before I was asking if you were working on the smaller-scale island, I wasn't sure whether to do all the work on the bigger version, knowing it would be for nothing.

But, it's never for nothing! I learned some things, some of which I spoke about in the livestream. So it's good that I made that test. It also reinforced the idea that everything is a bit too big.

Yes, but all of that is temporary work as I mentioned. There will be more work to break scenes into chunks, and optimise the ones that ones that are backdrops.

We really don't need 3 scenes loaded at a time. That also complicates things, because if you load a scene as a backdrop and later the same scene as gameplay, it means that when the scene is loaded as a backdrop it contains all the gameplay elements. Sure, you can disable them, but 1) they need to be loaded anyway and 2) even if they are disabled, those objects have a cost in memory.

Plus, having objects as prefabs in the gameplay scene allows us to bake GI at a different resolution (as noted here), while having them as scenes it would imply they'd come with their own full-res lightmaps, which would be a huuuge waste of video memory.

So again, for clarity, my plan currently is this:

[Beach scene]
- Beach ground
- Beach trees 1
- Beach trees 2
- Beach details
- Beach rocks
- Glade ground*
- Glade rocks front*
- Glade trees 1*
- Beach-unique gameplay elements (spawn points, etc.)

[Glade scene]
- Glade ground
- Glade trees 1
- Glade trees 2
- Glade details
- Glade rocks front
- Glade rocks back
- Beach ground*
- Beach rocks*
- Beach trees 1*
- Main island south coast 1*
- Main island south coast 2*
- Glade-unique gameplay elements

* = these prefabs are backdrop in this scene, so they are baked at lower GI resolution, plus have colliders and any gameplay script disabled.

We wouldn't load multiple scenes. You're either in Beach or Glade, and all necessary Prefabs get loaded as part of them, with the right settings.

 

I am indeed. I have carved up the small scale whiteboxing I put in the PR into scenes, following the plan that I had drawn (with small refinements). Playing with it I am not happy with dividing the town as it currently is. Maybe it is better to carve off the cooking festival (I guess here we will have many assets) into one scene and the rest of the town in one other scene since I can't find a convincing way to divide the town. What do you think?

 

We could totally have a scene-within-a-scene, they don't necessarily need to be one next to the other. So I'm imagining having the festival happen in a big square, surrounded by many buildings and accessible by 2-3 side streets, and the other town scene would be "around" it, being slightly bigger and a bit emptier, with less houses, less props, less characters.
What I can't figure out super well is how to put the main square in the centre of this scene without all its details. Like, how do you block the view? Tall houses? It would be a bit weird to have SO many tall buildings one next to the other in such a small town.

I'm imagining something like this (but without the fixed points of view). You can see how the square is surrounded by tall houses. But they do actually use a cubemap for the houses too! (old times...)

 

I am thinking about changing the map like this:

1. Beach
2. Glade
3. Fields
4. Cooking Festival
   
5. Town
6. Forest
7. Rocky Path

Banner over the entrance? posters? market stalls (backside)? oven? stage(backside)? carts? crates&barrels? A heap of water melons? fire wood? Some few trees(which would credibly stand there when there is no festival)? The only thing is all of that would not be very high (I guess we need to try whether that stuff would be high enough). From inside the festival it would be comparably easy - just the houses around the square and 2D behind .

BTW: I have made a script to write a rendertexture with alpha-channel from a camera to a png. So I place the camera from where I want the view and activate it if I want a to update the texture for the view (I am still thinking a 2D Image with transparent background is better than a skybox). The near plane of the camera can be used to show just the background.

 

I just want to chime in and say I'm loving the project so far.


I'm also especially enjoying the scale of the whitebox world in the livestream.
An 'empty' canvas (or an empty world, in this case) just begs for someone to draw on it!
So if the scale is reduced at all, _please_ don't go lower than 0.8 of the original 1:1 scale -- at most. I beg of you.

Why?

Exploration of real-world graphical problems (such as the location of the town center scene breakup in the suggested image above) is hugely important in the industry, and would suffer if our world was reduced any more than that -- plus, gameplay isn't strong enough to hold up without at least a _small_ sense of exploration -- even if that 'sense' comes from a mostly 'empty' world that just _seems_ full.

To add a point to this -- I noticed someone mentioned "Breath of the Wild" in the livestream chat when you went through the path in the Glade -- and suddenly the whole island appeared before you.
That's what they call a "wow moment" in the games industry -- and we need as many of these as possible.

The 'reward' for reaching the Glade was the breathtaking view of the grand size of the island -- and a promise of exciting adventure to the player. Even if there is very little _gameplay_ 'reward' in the game to be found in the game currently, keep in mind that there is a reason people are still playing Zelda OOT so many years later -- it was brilliantly designed (visually), with such limited resources.
Keep these "wow" moments and I'm positive more people would be willing to flesh it out with fun things to find and explore. For example, I know I am excited to be able to do so.

So, although I'm sorry to say I disagree with such an immense reduction in the scale by the amount seen in the video @Smurjo provided above -- I also see the value in a (slightly!) smaller play area, as per @cirocontinisio's suggestion. So I want to help find the right scale -- with something that gives us the best of both worlds.



Now, some tricks!!

As an oldschool gamer who loved Zelda OOT and paid very close attention to its design methodology to make the world feel much larger than it clearly is, a couple of interesting techniques come to mind:

1. If you pay close attention when traveling between the Temple of Time and Hyrule Market, you will notice that Zelda 64 uses a slower _white_ fade-out, compared to the _black_ transition it uses 99% of the time everywhere else. This simple (full-screen) transition is _immensely_ important for a number of reasons:
   • It sets the 'tone' for those sections of the world to be _different_ than the somber/quiet/reflective tone of the Temple of Time (and the rest of the game's darker themes).
     
     
   • The slightly 'jarring' white transition both feels appropriate (in the location / times it is used), and that very subtle 'jarring' actually distracts the player from the "hard" transition as the screen is wiped away with white and the user forgets where he was due to the huge contrast in tone from the Temple of Time to the hustle and bustle of the Castle Town marketplace.
2. In some cases, projected geometry on a flat surface (such as the foothills around Death Mountain, or the trees around Kakariko village before the Graveyard behind the Windmill) can solve a lot of problems -- but not as much anymore. Nowadays, that looks cheap and not very convincing, so we have to find a different method to this. Nowadays, we have what is referred to as "impostors" to work for the complex geometry like the houses / trees -- and "HLOD" to handle complex terrain geometry (like our beach scenes).
   
   https://forum.unity.com/threads/hlod-streaming-octahedral-impostors-using-megacity-subscenes.673390/
   
   "Imposters" are simply a shader trick. The material for these are generated by placing a camera in multiple locations at once and rendering each location to a different spot on a larger texture, then that texture is projected on to either an Octahedral sphere (or a half-sphere, depending on your vantage point being above and _not_ below or particularly straight-on), then these different portions of the texture/material are selected (based on your viewpoint to the camera) for display in the game.
   
   
   
   These are a little expensive on the CPU, but they give a lot of bang for your buck when rendering many instances of complex geometry (like rocks, trees, houses) on top of larger complex geometry (such as terrain around the town). See my thread here and see the (more technical) bit about Imposters HLOD too. HLOD has the benefit of displaying distant terrain -- and is fairly common today (with tools such as Simplygon that can generate HLOD meshes even in Unity).
   Two things are necessary to get this kind of rendering (that even works well on mobile! -- Fortnight uses it, so it is proven technology):
   
   
   • Rendering a camera in multiple places at once to achieve imposters, and with some simple math tricks and a shader lookup, you're set for the Imposters.
     
     
   • Running a tool like Simplygon on the complex mesh that is your terrain surface to generate the HLOD to display when far away from your terrain -- and swap that HLOD out with the legit, full-poly, mesh once you change scenes (and start to view it up close).

   ​

For us, cylinders might work for loading zones for something like 4 (in the new scaled-down world image by @Smurjo above) so that while you are inside the cylinder, you can get the "surrounding" geometry in the shot from all angles (as HLOD + Imposters), but when _leaving_ the cylinder, the dense area (in which you are standing) becomes the HLOD + Imposters, and the outer-area (5) becomes the "real" geometry (that can be explored):



To be clear -- I'm not talking about going with this picture's method of handling loading zones (i.e. separating 4 and 5), but rather, I am offering a clear way to handle such a use-case (assuming we go with this map as-is -- which I think is a great idea, assuming the scale is more along the lines of about 20% less -- at most -- than what it was before).



Outside of this -- I love love LOVE the approach @cirocontinisio uses on the Skybox and Reflection probe mechanism -- We _definitely_ need an Editor Tool to make those skyboxes easier to generate for more 'standard' locations than (4) and (5)!

I am only mentioning HLOD and Imposters because the technology isn't that hard to pull off (with a little math and shader know-how), but with @cirocontinisio's approach to generating a skybox for far away geometry via reflection-probes -- this keeps us from having to make _everything_ HLOD (and therefore bloating the game a lot), or _everything_ an Imposter (and keeping the game performant enough with highly unique objects -- such as nicely-dressed NPCs and uniquely-made vegetable stands -- that still look great too, whether far away, or nearby in the playable area!)


Hopefully this offers a solution that is not so "view-dependent" that it creates arguments amongst the community and Unity about the size/scale of the world, since it is would be using current, proven, (industry-standard) technology that, hopefully, is not out of reach of the math (and shader!) abilities of our community either!

 

Yes, but it's not something for this project
As much as I want to explore/showcase techniques to make bigger worlds possible, we need to stay realistic. Enabling BIG worlds is not just a matter of techniques used, but requires a lot of work from the team, assets, and a solid pipeline of production.

I do agree however that we shouldn't reduce it too much, but I just want to make sure we're aligned on intentions.

It might be cool to have a system to bake and display impostors. I wonder if we need it, actually. Except for the town within a town (which we're still discussing), we might never have the need to actually use impostors because you never get closer to the geometry or rotate around it. We might just do with planes with a texture (they'd still need to be baked though, and per-scene). Hmm.

For the same reason we probably don't need HLODs, but we might bake multiple meshes into one, yes. I think ProBuilder offers that ability out of the box.

 

@Smurjo I've merged your reduced map version, and I gave it a try.

I think it's a bit too small now? Everything feels a bit crammed, and as a result the verticality (which you've tried to keep, good) feels exaggerated too.

This is the current sizing:



Where the grey box is a 100x100 cube.

Consider that in the previous whitebox, the beach had a playable area of 100x100. The Glade was 100x100 (plus rocks), the space in front of the town, the town itself was more than that, the forest was also 100x100...

I like the bit where you come through the arch and see the town from above. I think having it a bit further away could work too.

I also like the size of the town, I think it feels right. But it feels weird it occupies the whole island almost.

Another thought: we needs ups and downs to be gentle, because often when jumping down a step or going down a steep slope, the camera temporarily goes underground. This is because the camera's collision behaviour has a delay. We can reduce it, but we will never be able to prevent it from going underground if the steps are high. So we need to keep it in mind.

I did a rough test and I scaled the map you made to 2, 1, 2 (so x2 but not on the Y):



It feels kind of right? Personally what I'd do is to give a bit more space to the forest area (6) so we can create a bit of a maze (forest is easy to fill, we have a lot of different plants), and make the mountain (7) taller. We can make the path that leads up to it gentle (that would be the playable area of the path) but the real peak would be quite higher, and unreachable. Or maybe we can make it reachable, but you wouldn't walk all the way up - you'd "teleport".

Regarding the fallen tree: I liked the canes It's also our only place where can use the Phoenix Chick's move to burn things (for now). Poor Phoenix Chick is pretty useless so far, haha.
In any case, the tree needs a round of refinement, it's not ready as it is and it looks quite different from the other trees. A bit too realistic too in the shape compared to the other nature we have. For instance the roots are too detailed.

 

@cirocontinisio I had already carved up the small map and was about to make a video, when your post about the different scale came. But I didn't need to be told twice. The small scale is doable but the bigger scale indeed leaves more room to be creative.

Not scaling up the vertical comes with problems though - I used it to restrict the paths where the player could move - now he can walk almost everywhere, even on top of the arch instead of walking through.

I reworked the whiteboxing to the new size, here comes a video about it

What do you see:

• view from the beach on to the water, beside Hamlet boat, we see a sunken boat with algae growing on it (in this clear water we could sink a complete Atlantis...)
• I am using the toon ground shader everywhere, but with the vertex painting and not that many vertices, I found I have limited control over the texture. Beside the 4 types of ground that were provided I made an endless texture for rock.
  
• I leave the fallen tree alone for now (albeit I like those roots). Given the reduced elevation I can simply walk past it.
• Then I walk through the stone arch of the glade and we get our first view of the town. Already in the small scale map I had increased the distance in which the town is seen, partly by rotating beach and glade and partly by streching the farms scene (it was and still is small to give the impression of countryside - we will have to make good on that by showing that there is something growing here)
• Then we walk into the farms scene and when we are a short distance inside the glade will be unloaded and replaced by some low poly representation of the glade which only has triangles on the front side. We turn and walk a short distance back to show that now the glade scene is loaded again
• Afterwards we walk further from the farms scene over the bridge into the town. On the bridge we get a nice view up the stream.
• We then walk through the town towards the montain path, on the way we cross the stream again and have look on a waterfall.
  
• Then we walk up the rocky path to the lake (I could have walked around that hill- but with the "flattening" you can walk over many things you couldn't before)
  
• Climbing the montain behind the lake we get a nice view from above, we see bit of the town, the farms and the arch of the glade. Only the town is really loaded - the farms are a 2D image and the glade is the "economy version" I described before
  
• The last part is about the jungle, starting from the towncenter. The jungle scene is not yet loaded, the few treetops we see are part of the town scene (I guess they are helpful to show the player where the jungle is)
• We have look from the jungle to the glade (and to the farms, but that's not done yet).
• Next we meet a plant critter, innocently hiding inbetween other vegetation
• Further on we barely see yet another waterfall - from the jungle to the ocean. There is a better view onto it from the other side, but I am not going there in this video.
• Then we cross the stream by a very special fallen tree. It is the art asset which we already had in the project, but with a tube shaped collider that I made for it. As you can see it is possible to steer the camera through it. But even if you don't - just press w and the collider will make sure Hamlet comes out on the other side.
• We are really getting into the shadowy atmosphere underneath the big buttress rooted trees now (It's possibly a bit too shadowy close to the player). According to my observation, with the camera under the tree crowns the trees have less impact on the view than the other smaller trees. I also think those roots are interesting to walk in-between (I have given them a mesh collider reduced to the minimum)
• Then we jump onto the island in the stream and back again
• Hiding (almost) under an overhanging cliff we find some slime critters. There is space enough for fighting them (but I don't want to show the combat system here - which worked fine). This is certainly a challenge to the camera, similar to the arch. I like the creepers hanging from the cliff and think of refining them some more (currently I am reusing a texture from a different art asset).
• Next we find some mushrooms. I wonder how the player finds out he can pick them (I have not worked on any picking system yet and wouldn't mind if someone else deals with it).
• I had to try out the bamboo, yet I don't find it very impressive.
• Last we get a nice view into the forest and beyond from an elevated position. We see there is still room for more maze, and if we need we can also expand the forest further into the ocean

Feedback and additional Ideas welcome. I also don't want give the impression I highjacked the complete whiteboxing - apart from the forest it is really not very detailed. I hope with the scenes kind of defined it is a good basis for others to refine whatever they like.

I guess I start on the cooking festival next. It will be quite a challenge to hide the fact that it is a different scene.

 

Super cool, I think in terms of size we are getting there! Did you submit it yet?

Don't worry too much about painting textures now. The ground needs to be sectioned anyway, in a way that makes sense to support breaking it into prefabs. As we do that we will add more vertices which will give more control over the texture, but first we need to settle on the scale.

I don't know exactly what you mean by "replaced by some low poly representation" and "loaded again". As I mentioned before, the plan is always to have 1 scene loaded at a time.

If there is anything that is too flat we can make it higher again. I just scaled the map to 2,1,2 to make a quick test, but this doesn't mean we can't have higher peaks or hills. We can also create instant unclimbable slopes with rocks.

I like this, we can come up with some fancy camera movement on rails for this!

We have what we call the "interaction system", pickup of objects is already supported (you must have seen it in some recent livestream).

Yes I think the forest will need a bit more space to evolve into a proper maze. I'm also imagining it a bit more crammed. Right now you look around and you can kinda see it all.

I had made this test which could be useful to show what I have in mind:

Obviously here there's no slopes or interesting points, just heaps of trees. But you can see how much more confusing it can be once you're in it.

A small feedback is that it's kind of hard to evaluate the shape of the terrain when there are big patches of black shadows. It would be nice if you could bake the lighting? You can set the light baking Texel per Units veeeery low (like even 1-2) to make it so the bake goes faster.

So yeah, if you feel happy with it, submit it as a PR so we can all take a look!
Thanks - great work as always!

 

I can't say I feel happy about it in the sense of it being perfect and it is still missing the cooking festival but I feel happy to share it, so I have opened a PR. To try out the map open the "SpawnScene" and press play in the editor. Make sure all other scenes in the folder were added to the build settings.

I baked the lighting. Somehow when I bake the lighting I always find something that needs to be improved urgently... so I ended up baking the lighting again and again. I think I also found the right darkness for the shadows now.

I understand that we will only have one scene loaded, but I didn't make the effort to create low cost representations of the adjacent scenes and teleport triggers etc. for the purpose of the whiteboxing. Instead there will be maximally 2 scenes loaded at a time using load and unload colliders with really minimalistic scripts which are not intented for the final version. When going to the teleporting we can still use the definition of the different scenes and we know which will be the adjacent scenes from which we partly need 3D representations.

We can also use the far away 2D views I have implemented. I made a small editor button for their update, you only need to open the camera for the view in the inspector and press the "capture view" - button to create an updated 2D view (make sure the scenes which you want in the view are loaded and deactivate the ocean). It still has the issue that the Texture2D which I create from a RenderTexture is too dark, which I suspect has to do with different pixel formats. I would appreciate help with this. I have tried post-processing on the camera, but unfortunately the post-processing seems to remove the alpha channel.

I think the view is deceiving here, these buttress root trees I made (and hope we will implement) are really gigantic (which is fully intended). I have started making a bit of maze a the lower entrance from the town, which should give you a better understanding about how much maze is possible (I guess some players will fall from the cliff when they run through the forest and see the cliff too late). Feel free to make more maze yourself in the other areas of the forest. I am not particularly attached to those bushes, I merely consider them a cheap way to represent "real" vegetation assets.

I also still have the issue that the waves of the waterfall always turn with the camera. How can I stop them from doing this?

 

Since the town is built on an irregular terrain like the side of a mountain, instead of having the houses themselves be really high, we can play with the terraforming and set the cooking festival in a place that's surrounded by tall rocks.
Just to give an idea:

 

@NicknEmart I absolutely love your solution. I was kinda thinking the same too, as a solution to block the views a bit. In combination with @Smurjo ‘s uptown – downtown suggestion, we could really split the scene in two (that "balcony" view could reveal the downtown part). Basically I always had something like that as a reference:


A combination of 2 actual places (above) that worked pretty well in the Helios map (below) in Overwatch. The part of the town you experience is flat and really compact, but gives you a good feel of a greater environment and a good Town-y/village-y feel. Of course not in that “mastodontic” scale ^^ I think we could settle with 15-20 houses for downtown/uptown and everything else in a skybox (pretty much how @NicknEmart has done). We could play with scale and distance to get the perspective right, I think it’s doable. And then you have a road for uptown which you’ll never cross but teleports you higher. So you‘ll get the nice views and the slopes without actually ever crossing them. (up-town and down-town areas could be in an almost flat surface).
That’s why I think teleport is more useful, I prefer that to a seamless scene. For example, @Smurjo in your medium size solution (which I think really works for what you are trying to do, but I am still missing some greater vistas, like what @NicknEmart enabled): I know it makes sense to cross the outskirts with the farms before reaching Town, but doesn’t really make sense to actually cross them gameplay wise, because I don’t think the game flow encourages any interaction with such an area at this point (I know I put them there in the first place, but I am rethinking about stuff xD).
I think of this like a montage, you have the film footages but now you have to juxtapose the clips in a convenient manner to bring your point across and keep the players attention. So after seeing the Town from afar and reaching that bridge you could instantly transition to the town center (the main square where the Main Festival takes place), instead of crossing a small environmental “filler”. Which does not really take time in your solution, but since it just contributes to the "atmosphere" atm, I do feel we could achieve much more if it was something bigger but in the horizon. I also always think about the sounds and the music. I would love it if after having a moment of peace and “wow” effect exiting the cave from the Glade you would transition into a loud - stuffed with sounds and smells/food - town scene (jumping straight to the action). Perhaps the main theme could still play there and slowly fade out. We could still keep the farms area for the visual connection or reach it later.

My point exactly, thinking like a director^^ I like the idea and I am really happy we have a Cinematics thread now. Taking this into consideration in the Quests&Scenario thread.

I am a bit surprised nobody has taken that task on before. I thought it was the most fun aspect of the game to think about! Are dense trees going to suffice for that? An easy solution would be to pick up a maze generator outcome from the internet and place the trees accordingly perhaps (cheap, I know). I really like the above environments you both made, don’t get me wrong, I am just missing an additional trial + error aspect of being in a maze, the pure level/puzzle design of it.

 

I'm not sure I understand. By "center" you mean the square where the festival takes place?

 

Yes. I'll edit it to make it clearer!

 

After @NicknEmart 's post I have separated off the cooking festival in a lower area indeed (it was the solution after I was kind of stuck with blocking the view on even ground) and it looks promising. I couldn't put it higher or else the arch on the glade would have to go even higher to keep the view. Also I didn't want to give a view on the jungle from above - except the one you get when you are in it. This is how the cooking festival terrain currently looks like (the colored part is the cooking festival, the town behind is grey).

I still have to block the nearby stuff from the view above (I am thinking with all kinds of sunshades and parasols, it's supposed to be tropical), then I will share.

There is absolutely potential for greater vistas - but they need to be made. In my PR I have a solution for generating the view images automatically once you have placed a camera from where you want the view (almost working - the automatically generated view is too dark, but I hope that can be solved without manual work soon). You see btw. the town from some part of the farms scene. This is using the same 2D as from the glade - which wouldn't work with a sky box.

Thinking about music you possibly want to give it a bit of time - from the view on the glade to to the bridge into town it takes you 45 seconds (30 when running) - that isn't much time to enjoy epic music including it's fading away and the slow start of the "town"-theme. I guess we might teleport through part of the farms scene - but we need something in-between, else we are not far enough away to get this large view. Given that usually most food is produced by farmers and we actually have kind of have a farmer character, we might possibly come up with some gameplay involving a farmer (who isn't very sociable but grows the best ... or the rare ... ). I like the general idea of not only getting recipies from different people but also getting ingredients from different regions.

I thought so too, yet I didn't want to spend much time on placing prefabs - not with the bigger scheme being so unfinished.

In my opinion there is enough "landscape" now in the jungle, for the rest all kinds of rocks and plants should do - except prehaps if someone has a very specific idea. Once my PR is merged feel free to play with it, there is so much space left.

 

Yeah, on a 2nd thought you are right. A good 30secs to enjoy it while running towards the town would be nice, before any transition. And I liked that farmer too. I will wait to try out your scene before trying new stuff out, just wanted to give my thoughts for the town mainly. But I like how this is evolving, really helps to see the environments forming!
EDIT: Should the Forest Maze become a separate task to encourage more contributions? At this point I wouldn't take the initiative to open a new thread but I think it would be fun if people started to come up with maze ideas. I know the clock is ticking

 

I guess it's time to share what I have been up to in last days.

Remarks:

• I have not bothered much with scene management. I hope I can take over @cirocontinisio 's solution with Addressables once it's ready.
• I wonder whether I should make a separate pull request for the art assets (would that speed things up?) Currently there is no rope joining the bamboo frames albeit I build them in a way that one can imagine how rope would hold them together. Shall I add it or not? I fear with rope the triangle count could double easily.
• I am using the detail mask tint shader for the sunshades and parasols, this way they can be made different colors and even different patterns. Yet the detail mask shader has no wind - which would be quite nice on the cloth items, as it can be seen on the banner. Also a double sided shader would be good as it would save half the triangles with the cloth items.
• I have made 3 entrances to the cooking festival. One of them is a steep, dark and narrow "secret path". Here the camera could do with some help if we implement it.
  
• I also found that the upper part of the houses is bigger than their collider. If you stand beside a house and turn, the camera goes inside the house (the house on the opposite side of straight path down to the cooking festival is especially liable to that as you would naturally turn to go down the path). Should I change the prefab? It was made by @laczkyl. Or should I put an additional invisible collider at the complete row of houses?
  
• I have provided 2 spots for views from high up albeit right now there is still the real scene to be seen.
  
• I fear the particle systems are kind of costly but I totally like that steamy cooking atmosphere they make. It would also be nice if we could keep at least one particle system to be seen in the far away view.
• Obviously the cooking festival poster is just a placeholder. I guess it would also have to be localized the same as the "welcome" banner (if we go with poster and banner, we could also make a banner saying cooking festival).
• There still a lot of room for ideas and art assets in the cooking festival scene

 

So I tried to play with this Town within a town thing a bit more based on the above contributions. I did introduce some Big Rocks and Trees as those prefab walls and came up with something like this (don’t mind the little details and the ProBuilder meshes I only wanted to do a basic layout):

Some parts could be optional as I mention in the video. Is this what we had in mind?

I am really wondering how heavy the final scene might be and if the loading/unloading parts could actually help the way I imagined it (trigger boxes are really rough,we 'd need better custom shapes). I guess we will see in the end, but it was fun trying to figure out a solution. Also wondering how this might work with Addressables (it was a lot of new info for me that’s still sinking in to fully think in terms of them when designing) but I’ll come back to it, it was well explained.

Ι am also curious if/how we might implement any impostor approach as discussed above regardless of our case here. In what sense is it different to the low-res prefab backdrops that were mentioned and where could these texture planes be used in this example? (I guess any huge prefab in the distance that’s not part of the skybox could be a texture plane? Something far away enough that you wouldn’t notice its 2D while moving?)

Keen to hear your thoughts about it and willing to improve any parts before submitting a PR.
PS1: The scene was build on top of @Smurjo ’s Town Ground Mesh on her MediumSizedWhiteboxing and was inspired by her solution and @NicknEmart 's approach with the vision blocks and the height differences (plus the references I shared in my previous post). Thinking the scale might still be a bit big..but easily adjustable. Also introduced some stairs, that might be converted into ramps, just to give some extra sense of structure.

PS2: Didn’t place any Key Character yet I’d like someone else to place them in the scene as they’d see fit. I placed Fry King at some point but just as an example. I was actually thinking of placing the Farmer there and have the Upper Part be the place for the crops/farming to be done instead of the Town Entrance. The idea (despite the cool view) came from the Quest and Scenario post where we discussed about the necessity to find the Farmer after obtaining the Rock Critter eggs to have him take care of them and multiply them. So I guess they’d need a Mountainous environment to live and thus the Upper Part could serve as a continuation of the mountain/ transition area between the Rocky Path and the Town.

PS3: I think that by adding the other scenes in the Skybox to fill in the horizon and also playing around with this stretching illusion , we could totally make everything appear like an open world without it being open world at all. We don’t have to walk everything, I think the trick is just to see what lies in the distance without realizing where the walkable area ends and the skybox/texture planes start.

 

It's great. It feels like a town without actually having many houses. The fish market could be a separate scene and the uptown too.

 

I agree with @Smurjo, it definitely feels like a lively town even without many NPCs in the whiteboxing, great job! Thinking there might be more NPCs outside each house and some behind the stands as vendors, possibly with more in front trying to buy stuff.

Based on this Quest and Scenario post clarifying the key characters, is it just Gourmando (saucier) and the Baker that would actually be in the town since Bard Hare is first met on the Beach and the Desserter is in the Forest Maze? If so, here's some thoughts:

• Yes, the Farmer should be up near the outskirts of town, as the farmland isn't going to be in the town itself, and the mountainous ideas make sense to me
• The Baker should be near the oven that's near the path to the secondary square and have one of the stands a little closer to the oven so they can bake and present goods at the same time
• Torn between somewhere in the secondary square and the fish market for Gourmando, not sure the correlation of either with his job, just a gut feeling
• Should Fry King be greeting the player upon entering the main festival area, or some NPC, or no one?
• Is that a well in the middle of the main festival area, or just a table? Whatever it is, maybe Fry King can be sitting in a chair near it, waiting to be impressed? This would allow a little bit of the final victory song idea I wanted to try since the spectators could all be looking into the center, maybe chanting something in anticipation, or having Bard Hare singing to narrate that Hamlet is about to present his dish, pausing for judgment, then giving a victory cheer

I'd be interested to try and play around with the whiteboxing to do these things, any suggestions on the best approach? Since I'm new to using these tools, I'm assuming placing the NPCs is part of scene loading/unloading, right?

 

Wow, this is really cool! The layout feels very organic and plausible, full of interesting spots to discover. Really good job!

I will repeat it once again because it seems like this confusion will never go away: we don't load/unload parts of the scene, we only load entire scenes. So whenever you go through an arch or down the stairs or up the stairs, you'd see a brief fade to black > the new scene will load, taking as much time as needed > fade in from black and you're back in the gameplay. That's why we need the "prefab walls" they block the view so you don't see that the rest of the island is not currently loaded.

You need impostors when you have something in the distance that you don't want to load as a 3D model, and you can move around it. Then having impostors prevents it from looking like a 2D plane.
But we don't have impostor technology right now and I don't foresee we will have one, so I'd like to keep things simple and rely on close 3D prefab walls + far 2D planes + skybox.

Honestly I think it's pretty interesting, and it would be nice if you could submit it.
If anything, I'm just slightly worried about the parts up top: when you're there, you see the whole island below, so we will need to be quite careful with prefabs and basically the end result is that you will see a very empty island and town. So no characters, no food, no grass or small details, but only the bigger trees, houses, and food stalls. It might look a bit 'meh', but we'll see.

Still not too sure about that, you only rotate the camera horizontally in your video and I think if you do it on the vertical axis, the illusion will be broken revealing the lack of connection.

 

They would be placed in the scene, so they are part of it and they are loaded and unloaded when you enter/exit a location.

 

Thank you for the feedback!

Ah my bad, I got confused indeed. But at least the same approach could apply to the different scenes loading as well I guess.

We could always place some trees/rocks etc at the end of the prefab "connection" to hide any irregularities, or just use this only in a horizontal movement then. Not sure if this could really work either, def needs more trial and error.
Thanks again for taking the time to answer! I'll go ahead and submit a PR then, but before I do it:
1. Should I do it on the Main Branch or the Art Assets branch? I see the Art Assets one has the Whiteboxing folder, that's why I got confused.
2. I haven't baked a lightmap yet and if I try it shows me an estimation time of 2h Is this normal? I already started it so I'll try to include it once its finished.
3. I just wanted to clarify on the setup that one needs to do to play a new scene, because currently dropping the EditorInitializer (+CharacterSpawner+CameraSystem) didn't let me navigate around, I saw that you need to give it a LocationSO to work properly and be able to navigate around with Hamlet.
Is that correct?
So what I did was go into Create > Scene Data>Location and there I made a LocationSO, where I've dragged and dropped my scene in the "SceneReference" slot (which also made it pop up in the SceneSelector tool). Is this OK?
If it is, should I include that SceneData file in the PR as well? (they are all placed in a Town Folder inside the Whiteboxing Folder).

 

You can do it on

main

, we dismissed the

whiteboxing

branch. And yes, place any asset in a folder inside the whiteboxing folder. Thanks!

Sounds a bit much. Maybe your Texel per Unit setting is a bit too high (40?), you can start really low if your intention is only to get a perception of the lighting. Like even 2-5.
Also make sure you're using GPU Lightmapper, not CPU.

Yes, true, you need to create a SceneDataSO for the scene you want to use now... I should add it to the wiki (sigh).
You can submit it as part of the PR, even in the regular SO folder (not in whiteboxing) as an exception. Just name it so we know what it is.

 

Ok thanks it was reduced to 4-5 minutes. Also took me a while to realize the Directional Light has to be set to "Mixed" otherwise the characters would appear black (just putting it out there, some things are not always obvious to all). I saw that on the other scenes you have used LightProbes, is this something you'd like to see in a PR as well (in an ideal world)? I've seen one of your talks on that before, I never played around with them yet but would like to try, I just didn't have the time rn.
Anyways, you can find the PR here: https://github.com/UnityTechnologies/open-project-1/pull/384

 

Thanks!

Generally we need LightProbes in the game, but you can also just put one in the sky when you bake. The character won't reflect light changes when it walks into shadowed areas, but it's fine. We can place them correctly once we have the final layout, since it's a time consuming operation.

 

Another stroll in @itsLevi0sa's whitebox, now with baked lighting, and the town music (surprise). Looks nice, eh?

Some thoughts:
- For the forest, we don't need to put it all on the big map. Could be just a bunch of trees, like a very thick but not huge grove. But then, when you enter the area, it's huge and feels like a maze.

- Another idea I had for the forest would be to force the camera from the top, above the trees, and you wouldn't be able to bring it down. Something like this:

But with many more trees, of course, so that you'd barely see the player and where you're going. And with no points of reference, it would really feel like a maze. We could have the occasional lack of trees, an opening, so the player can orient themselves.
Good or bad idea?

- What should we do when the player is up high, and could jump down into the town? Should we inhibit the jump? Should we put an invisible wall so they can't jump down? This could look bad, and easily spiral so that we need to handle a million invisible colliders. On the other hand, if we allow jumping we need to handle the colliders of the rocks, impede the player from getting stuck, and most importantly make sure they can't walk off the map. Thoughts?

 

I would think it's a bug if I can't control the camera the same way as in the other scenes.

Did you see I have updated the trees?( https://forum.unity.com/threads/3d-art-assets.980892/page-4#post-6873101). As I wrote there I would like to create this shadowy atmosphere with sunny pathes in the forest. You wouldn't get that if you stay above the trees with the camera. I fear since you removed the trees you had no chance of testing it (actually the forest is no forest without trees).

 

What a nice surprise! Double with your own take of the town. I find it really interesting to see someone else experiencing the setting, seeing what decisions one makes, where they stop, what got their attention, where they slow down their pace. That 's the best feedback. The music is great! It's so strange how consistent all sounds are with their scene yet with one another too, well done to the artist (Marta Ascari!). It's really a great feeling seeing everything come together and I am glad we got a good timing on this (thanks to the livestream delay)

I personally would put colliders to not let the player fall, why does it mean we'd need millions of them? Wouldn't a custom shape along each way suffice? Also since the Upper Part is going to be a different scene I guess the player would witness the black fade in/out anyways if he'd try to "jump" on that connecting path. (I know its pretty steep, don't know how we could avoid the camera going through objects at this point, so I guess the loading/fading could skip that part).

I love the force top-down camera view, introducing a new style of gameplay for a while. I like it every time I experience that in a game actually, thinking especially of Nier Automata. It makes total sense for that specific part. I think your approach would make the player feel lost indeed but how could we give them hints to find their way? This approach would need some landmarks so that the players can rely on observation to tackle the challenge. Or they could "drop" markers along their way to help themselves. (I know that's a new mechanic that we don't currently have and that I have already asked if its possible for the Mountain). I was also wondering, especially after the Quests and Scenario diagram that Chema posted if we are actually going to use our Phoenix Chick fire to burn obstacles down or not. Because in this case, perhaps we might place some burnable obstacles in the maze and that's how the player might find the center of it.
In general, I'd personally like the player to observe some kind of pattern that we thought of. Both in the forest (spatially) or in the mountain (with the critter behavior), but maybe we won't for now and that's ok. I know we have roughly a month, so I'd really like to make a thread just for the Maze and discuss all this, have some game design chit chat, I think it's a really interesting topic that anyone could post references regardless of what we'll implement in the end.
To conclude: I think the top down approach with an intro and outro from @Smurjo 's forest take would be ideal cause I really liked the environment she showed on the video above, either as the one we'd find the Desserter in or the one we'd experience before entering the Maze and switch to the Top Down view.
Either way, I trust you'll/we'll come up with a nice solution! I've player one of your shipped games (UFHO2) and the spatial puzzle was strong in this one, I really enjoyed it. So now I know you like and worked on interesting brainteasers before regardless of what we might manage to accomplish here, it's the best opportunity to get to know who you "work" with! Needless to say the music from the artist on that one was great too (Fransesco D'Andrea).
Would do the same for any other collaborator here who might have shipped a game.

 

Nah, there will be plenty of situation where camera control will be taken away, like in narrow tunnels and other situations where it's good if the camera is on rails.
To make it extra clear we might also display a symbol (like, a camera with a forbidden sign, or something...) when the player tries to move the camera in these situations?

In this case though I would use the mid-size trees we have already, so that the camera can sit on top of them without being too far from the action below.

I saw them, sorry for not giving feedback. It seems to me that they still need some work, both in the modelling and the texture, which right now has a lot of distortion, and lacks some detail especially on the trunks.
Plus, I find them VERY tall. A bit weird compared to other trees we have, and I know trees exist in different sizes in nature, but they feel a bit extreme as of now?

 

That's a good idea. So we can have one big collider to catch the player falling, and when they do, they get teleported to the scene below.
I don't like the idea of putting invisible walls to prevent from jumping anyway.

 

https://github.com/xraxra/IMP

That's not doable for us? -- I mean... the technology is already there... and it seems pretty straightforward to translate into URP (nothing fancy -- only a bit of uv math and a texture made from simultaneous camera input).

This is why I suggested imposters for reasonably large / complex scenes. This way the colliders you have to deal with are only the ones you actually need for actual traversal -- and nothing more.

In cases like this, where we need the environment to feel "explorable", it looks like we'd need imposters for the large number of rocks we're using to hide any potential distant scenery we might be able to hop down upon. Closing everything off makes the world feel like a dungeon crawler -- and pretty claustrophobic -- overall.

I'm kind of torn on this "transition-based" approach.

While, in theory, it would 'work' -- it would always be jarring to the player for the whole screen to suddenly fade to black despite the player being able to clearly see the ground beneath them.

If you look at Zelda OOT (which seems to be the kind of old-school methodology we're going for here), when Link falls from a high floor in a dungeon, the upper-floors fade to black (vertex coloring) and are unloaded from memory, and then the lower-floor fades in (from black), with Link being on-screen all at once. These aren't new "scenes" but they're just clipping areas where other scenes / objects are patched in, reusing old pointers wherever possible, and resetting the room to take into account the newer areas during the falling transition.
A hard fade-to-black and return-from-black here would have been extremely jarring when trying to traverse dungeons such as Jabu-Jabu's belly or areas like in Majora's Mask's dungeons. The pace of the game would have felt like a slog.

The difference between Zelda OOT dungeons and this town area is that going from "floor-to-floor" in Zelda's worlds is all about clipping-bounds and object-pointers.

This meant you could hide the previous scene with the black ceiling as the new 'scene' (or rather new 'clipping-bounds' and object-pointers) were being loaded in realtime during the fall transition (as you neither saw the upper-floor or the entirety of the lower-floors of the dungeon as you fell -- yet neither did you see any ugly "fade-to-black" loading screen when traversing from floor-to-floor in those dungeons (or in the overworld) either (as this would kill the game's pacing.).
To save pacing (and a jarring scene transition) is not going to be easy to do (or technically possible) while maintaining a modern graphical level-of-detail as well as a proper pacing of exploration (at least on a graphical level) without some technology behind us to enable that.



That said, we need something akin to either Imposters, or some kind of tooling to allow easier clipping-boundaries, LODs, and/or view distance -- or turn the game into a dungeon-crawler (or have an extremely fixed / limited camera).


This, I suppose, is up to you, @cirocontinisio. You're the man.

We can prototype it, as always -- but I'm positive you're going to find it as jarring as I do at some point.

 

@cirocontinisio regarding looking down at the town from up top, can we cover the whole festival with big circus like tents. That way we will see the tents from above and won't need to worry about the empty town.

We can also do tree jutting out from the side of the mountains to block view and defining player bounds too.

One thing I noticed the houses are looking a little awkward sitting flat on the uneven ground. Can we give the houses a circular or square base for foundation? What do you think?

 

It depends, it needs to be tested in our project. We're using a custom shader, so I wonder if this tool then allows you to bake lighting or... how does it get it once you use the impostor prefab in place of the original one? (maybe it uses light probes?)

I was talking only about the top parts of town, they are maybe 3 paths all in all.

I think you're overthinking it a bit. There are many games which transition between scenes or falls by unloding and reloading the whole game world. Actually I would find it weird to see the character against a black background as the game transitions between scenes. A full transition to black is much nicer.

Also if we don't want to do that, we first need to rethink the loading/spawn system to allow loading/unloading scenes while conserving the player Prefab in the scene. I don't see how this rework is really necessary.

 

Might it be a solution to place a few an animated 2D sprites of talking or shopping townsfolk in front of a 2D-Image to create a credible view from specific spots higher up (and block the view everywhere else)?