Revit, Blender & VX

[McKeu]

Heyas,

As my fiancee is an architect and mostly works in Revit, and I know that export options from Revit to Blender exist (http://wedesignvirtual.com/preparing-a-revit-model-to-export-to-blender-3d/), has anybody any experience how the results are in conjunction with VX?

Is a lot of afterwork required?

Any input would be appreciated, as this could totally increase our room (or literally anything) creation speed.

Cheers.

[MarkZappa]

Well, it is and could not be a hit at the same time if you know what I mean, you'll have a mesh in principle but you'll probably won't have the proper mesh for a game engine just with a touch of an import-export simple manneuver

Most of these CAD software heavy industrialy inspired like Autodesk Revit, Sketch Up, SolidWorks, Alias Autostudio or CATIA have an unusual constructive workflow that would never match with video game industry, namely you can easily create a panton chair in SolidWorks with a set of three nurbs on three parallel planes coating it connecting the dots while you add thickness to the resulting surface and voilá, it's alive but... import that in Blender and you'll have a geometry nightmare and that's just a simple chair

These workflow tends to generate n-gons or meshes that would require a lot of work, not mentioning that vision of architects or designers is conceptual and schematic which once translated to a videogame, no matter how much quality H5 can deliver it will look death, dull and poor
Further more on these softwares workflow, is like boolean workflow in Blender, endless possibilities and a nice trip but pray for the average game engines to cope with that unsolvable geometry without a full retopology first, and to make myself even clear... it would be better to create the whole mesh from scratch than work in solving the imported object

Now, about the second part, the only way to accelerate the room creation process is counting with more people porting rooms, and is not that there is no resources in that area, there are plenty of DAZ rooms to be converted, but looking further deviantart XNA groups offers a lot of game scenes rips

I'm not saying is not possible but I want to remark it would lead to nightmarish ports and back ache

[Sexvision]

I thinking more of making some standard objects like modular walls to make gains in creation speed. with some standard parts we could recreate rooms like in sketch up within hook and give everyone creative freedom. Walls, doors and windows are easy to make in blender.

Im also working on a database with hook objects to give everyone the chance to make there own creations. this will be my final goal and the reason why i port my rooms modular.

I port my rooms from honeyselect 1 as they are already optimized as its also a open gl game. Adding the norm and specmaps makes it perfect for hook 5.

Pm me if you need some meshes to test. i will give you something to do

[x17]

Regardless of from where the mesh is imported, there are several things to check and fix if necessary ...

- n-gons (polys with more than 4 vertices) - they should be turned into either quads or tris (4-sided or 3-sided). Its not that the game is incapable to run with n-gons, but the n-gons require additional script checking and fixing, and aside that, theyre bigger performance hog. So, its really pointless. We can triangulate whole mesh or parts of mesh in Blender, or we can use CtK synthax to triangulate. If scripts within the game asset affected with n-gon arent corrected, the game wont display the mesh at all. Even the log of CtK will show this and point out to n-gon. So. n-gons are generally pointless here ...

- removing doubles / merge by distance - fixing of this is MORE IMPORTANT than it seems to be. Firstly because of potential CTD that can occur if verts are overlapping too much within the game engine, secondly because of clearing of excess and pointless vertices. Third, because this usually fixes problems with broken mesh when posing (sometimes) and is neccessary anyways.

- fixing normals - this should be always done in the ending. When all mesh work is done. The normals are recalculated to the outside. Please note that if youre converting with any Blender that is above 2.79 version, it will require fixing the scripts by adding few lines on specific places (you can check my modding tutorial section input for that), but all is doable and there are no problems. Overall, CtK was made for usage with Blender 2.64-2.79, but its not problem to use it with any new version.

 

 

[McKeu]

Thanks for the input guys.

I think I will let her whip up some simple structure with and without textures (materials) in order to see what happens with both import methods (blender plugin for ifc files, and export as fbx). I guess it could be interesting to see, as Revit comes with all kinds of prefabs to put in your structure and we do have an industrial sized database of prefabs at our disposal. Just need to see how Blender deals with Revit furniture, materials and such.

[McKeu]

And just for comparison I will probably try to import one of her more complex works, like this supermarket and watch my PC go up in smoke 

[x17]

Hmm, interesting projects. Overall, from a standpoint of TK17, there are several things ...

1. Try to keep polycount manageable - especially when its about human-made objects and buildings. Vegetation is problematic to represent realistically with its relative high polycount need. Its mainly because within this game engine, its not made for large open spaces and has no modular, partial loading. Like its, for example, case with newer modern game engines, especially something like UE.

The polycount of our scenes within TK17 should not go over 1 million vertices per room in any case. Still, we should keep it reduced and optimized whenever possible, so keeping it around 700-800k max is ideal. Its not only the room that gets loaded, its also bodies and every clothing asset. The rooms are actually one of parts that can be optimized to very large extents. Especially when its about buildings and houses.

2. Texture drawcalls is another thing - trying to keep textures as low as possible is a thing within rooms. Especially for rooms that have many textures by default. Its smart to separate textures in logical way - like for example, have one presumed material, glass and transparent-like materials on one texture, have concrete or plaster on another... etc. Its to mainly easily utilize transparency if needed.

This texture merging and remaking of UVmaps requires good time and effort. But will optimize the room significantly. Otherwise, the room could be unplayable and very unstable if there are many drawcalls.

3. Its important to use certain tricks - utilizing a HDRI dome for example to get a certain illusion of space. The places that arent intended to be visible or be used with our models dont need to have any large detail and can be very crude like. Its important only for that , what should be visible and where models would pose in, to be in sufficient detail. For majority of stuff really big textures, like 4k are truly unneccessary and are "overkill". It all depends. Sometimes we need that extra detail and 4k could give an edge. 

[McKeu]

25 minutes ago, x17 said:

Hmm, interesting projects. Overall, from a standpoint of TK17, there are several things ...

1. Try to keep polycount manageable - especially when its about human-made objects and buildings. Vegetation is problematic to represent realistically with its relative high polycount need. Its mainly because within this game engine, its not made for large open spaces and has no modular, partial loading. Like its, for example, case with newer modern game engines, especially something like UE.

The polycount of our scenes within TK17 should not go over 1 million vertices per room in any case. Still, we should keep it reduced and optimized whenever possible, so keeping it around 700-800k max is ideal. Its not only the room that gets loaded, its also bodies and every clothing asset. The rooms are actually one of parts that can be optimized to very large extents. Especially when its about buildings and houses.

2. Texture drawcalls is another thing - trying to keep textures as low as possible is a thing within rooms. Especially for rooms that have many textures by default. Its smart to separate textures in logical way - like for example, have one presumed material, glass and transparent-like materials on one texture, have concrete or plaster on another... etc. Its to mainly easily utilize transparency if needed.

This texture merging and remaking of UVmaps requires good time and effort. But will optimize the room significantly. Otherwise, the room could be unplayable and very unstable if there are many drawcalls.

3. Its important to use certain tricks - utilizing a HDRI dome for example to get a certain illusion of space. The places that arent intended to be visible or be used with our models dont need to have any large detail and can be very crude like. Its important only for that , what should be visible and where models would pose in, to be in sufficient detail. For majority of stuff really big textures, like 4k are truly unneccessary and are "overkill". It all depends. Sometimes we need that extra detail and 4k could give an edge. 

I think from our (my gf's and mine) perspective optimizing a room inside Revit before exporting in regards to polycount, etc. might be simpler, as she has high experience in dealing with that and in Blender we both have basically zero

Then I'd need to learn optimizing it inside blender.

[x17]

1 minute ago, McKeu said:

I think from our (my gf's and mine) perspective optimizing a room inside Revit before exporting in regards to polycount, etc. might be simpler, as she has high experience in dealing with that and in Blender we both have basically zero

Then I'd need to learn optimizing it inside blender.

Overall, as long as you could import the rooms from Revit into Blender without losing anything, its OK. The buildings and similar objects are possible for high optimization - many walls can be deduced to only one single quad, or just few polys. In general, walls and similar surface really do not need any large diffuse textures. For details, though, we could use normal maps sometimes. Like for certain plaster or wooden surfaces. Or brick walls, or similar. It goes fairly into room section, too, that normal maps are the ones that should be in higher res when needed. While base diffuse and specular map can be small.

It all depends on what is done, ofc. 

The vegetation and nature are definitely more trickly in game. Plants would quickly add onto total polycount.

The most work in rooms, though, is always around rework of UVmaps and above lowering or number of texture to reduce drawcalls.

[McKeu]

1 hour ago, x17 said:

Overall, as long as you could import the rooms from Revit into Blender without losing anything, its OK. The buildings and similar objects are possible for high optimization - many walls can be deduced to only one single quad, or just few polys. In general, walls and similar surface really do not need any large diffuse textures. For details, though, we could use normal maps sometimes. Like for certain plaster or wooden surfaces. Or brick walls, or similar. It goes fairly into room section, too, that normal maps are the ones that should be in higher res when needed. While base diffuse and specular map can be small.

It all depends on what is done, ofc. 

The vegetation and nature are definitely more trickly in game. Plants would quickly add onto total polycount.

The most work in rooms, though, is always around rework of UVmaps and above lowering or number of texture to reduce drawcalls.

I mean, in above supermarket example you could simply lose the upper floor, as it is unnecessary. That would also remove the overkill lamp installation. Also the outside steps and front area can go easily, as it's an inside area room only. You'd also need to lose the logo at the entrance, obviously. The wooden hanging installation also is not necessary.

Then reduce it down to as little polys as you can, and that should do it, I guess.

[Sexvision]

4 hours ago, McKeu said:

And just for comparison I will probably try to import one of her more complex works, like this supermarket and watch my PC go up in smoke 

Please film that when you do. I like PC with flames

[Sexvision]

I have noticed that round objects have a lot of polygons that is killing framerate.

for example the lights in the main hall will be very heavy. try to keep them as a joined mesh will gain you some fps. Round objects in general are fps killers.

Also the the ceiling that is brown seems like it would have a lot of detail. there is a lot to load there.

You can join meshes but will need to do a retexture after that. for that you may want to look in to texture painting in blender 3.0 that will make things easier.

I normaly seperate meshes on the count of what specular map they will have. this will give you more control on setting the right spec values in the game without changing things you dont want changed.

So everything thats metal as a separate mesh, and all lights that i would set a glow spec on would be a single mesh. i can then push the values of that paticular mesh.

This will work for h5m rooms the best but will also give you a separate codeline in a addon vx room that you can manipulate with shine and emission.

I also use normals to create bricks. you can use any good normal map to do so. (hook 5 can handle normal and displacement combined to give a better result)

You can redo the logo in blender very easy by using text and change them to a mesh after your achieved the look you want. you can control the poly count in text easy and put a glow specmap on it to make it shine.

as a hook user i like to make walls and beams low poly because it will let me use the scale option in hook and i can manipulate the objects in game. High poly items will not let me do that and will become a fixed object in size.

 

[McKeu]

2 minutes ago, Sexvision said:

I have noticed that round objects have a lot of polygons that is killing framerate.

for example the lights in the main hall will be very heavy. try to keep them as a joined mesh will gain you some fps. Round objects in general are fps killers.

Also the the ceiling that is brown seems like it would have a lot of detail. there is a lot to load there.

You can join meshes but will need to do a retexture after that. for that you may want to look in to texture painting in blender 3.0 that will make things easier.

I normaly seperate meshes on the count of what specular map they will have. this will give you more control on setting the right spec values in the game without changing things you dont want changed.

So everything thats metal as a separate mesh, and all lights that i would set a glow spec on would be a single mesh. i can then push the values of that paticular mesh.

This will work for h5m rooms the best but will also give you a separate codeline in a addon vx room that you can manipulate with shine and emission.

I also use normals to create bricks. you can use any good normal map to do so. (hook 5 can handle normal and displacement combined to give a better result)

You can redo the logo in blender very easy by using text and change them to a mesh after your achieved the look you want. you can control the poly count in text easy and put a glow specmap on it to make it shine.

as a hook user i like to make walls and beams low poly because it will let me use the scale option in hook and i can manipulate the objects in game. High poly items will not let me do that and will become a fixed object in size.

 

Yeah, I mentioned some of those points above:
"I mean, in above supermarket example you could simply lose the upper floor, as it is unnecessary. That would also remove the overkill lamp installation. Also the outside steps and front area can go easily, as it's an inside area room only. You'd also need to lose the logo at the entrance, obviously. The wooden hanging installation also is not necessary.

Then reduce it down to as little polys as you can, and that should do it, I guess."

How to finally do that in Revit is not my task

[Sexvision]

16 minutes ago, McKeu said:

Yeah, I mentioned some of those points above:
"I mean, in above supermarket example you could simply lose the upper floor, as it is unnecessary. That would also remove the overkill lamp installation. Also the outside steps and front area can go easily, as it's an inside area room only. You'd also need to lose the logo at the entrance, obviously. The wooden hanging installation also is not necessary.

Then reduce it down to as little polys as you can, and that should do it, I guess."

How to finally do that in Revit is not my task

I would say port it and see what it does before getting all the detail out what makes your girls supermarket look great. Set a goal on where you would like to be fps wise and work to that.

Keep in mind that every person you add to your scene will take about 25 fps so also test it with 4 models to see a true fps count.

For blender just learn what you need. there is no point in learning all of blender for your goals.

[McKeu]

11 minutes ago, Sexvision said:

I would say port it and see what it does before getting all the detail out what makes your girls supermarket look great. Set a goal on where you would like to be fps wise and work to that.

Keep in mind that every person you add to your scene will take about 25 fps so also test it with 4 models to see a true fps count.

For blender just learn what you need. there is no point in learning all of blender for your goals.

I will have to change it a bit, as this is a real life supermarket that was actually built here in Bucharest

Can't have that in a game like this.

[McKeu]

So, if we take an existing project in Revit, export it as FBX, which seems to be more compatible with gaming than IFC, it ends up looking like this, which is a promising start. For basic structure importing this seems like a great way. It even kept furniture, families and shit.

I used the Revit included example structure, as I cannot make any projects of my girl public. But she can whip up a structure in 20 mins flat.

Now, I assume you would have to import a character to actually adjust the scaling of the building, eh? Or would it actually take the original measurements (Revit uses both metric & imperial, while blender counts in units responding to like 1 meter, right?)

But, tbh, the controls in Blender are just fucked up. Wanted to move the camera instinctively with right mouse and ended up moving a wall around 

 

[McKeu]

So the very least we can achieve at this moment is to provide basic buildings for others to work with. As my knowledge with Blender grows, probably even more.

If anyone with some Blender experience could look at attached file (Extract, and then import as FBX) and tell me what the amount of afterwork required would be (not do it, just assess, ignoring outside terrain), that would be absolutely grand. That would tell us what prework in Revit would already help.

Maybe @Sexvision or @sadekhnd?

rac_basic_sample_project-3DView-3D.rar

[sadekhnd]

1 hour ago, McKeu said:

So the very least we can achieve at this moment is to provide basic buildings for others to work with. As my knowledge with Blender grows, probably even more.

If anyone with some Blender experience could look at attached file (Extract, and then import as FBX) and tell me what the amount of afterwork required would be (not do it, just assess, ignoring outside terrain), that would be absolutely grand. That would tell us what prework in Revit would already help.

Maybe @Sexvision or @sadekhnd?

rac_basic_sample_project-3DView-3D.rar 5.63 MB · 0 downloads

What room, h5 object room?

I would group stuff together. Building alone and room stuff together and doors  (you can open/close them in game).  It has 257 k  faces, so it's not that much, It will work smooth. Many stuff like seats are the same so you need only one and you can clone it later in game. (better performance).

Overall the room looks cool. Terrain auch.

It will much work with textures. I would use same textures few times, like one texture for few walls or one kind of leather for seats.

Thats all by now  

[McKeu]

36 minutes ago, sadekhnd said:

What room, h5 object room?

I would group stuff together. Building alone and room stuff together and doors  (you can open/close them in game).  It has 257 k  faces, so it's not that much, It will work smooth. Many stuff like seats are the same so you need only one and you can clone it later in game. (better performance).

Overall the room looks cool. Terrain auch.

It will much work with textures. I would use same textures few times, like one texture for few walls or one kind of leather for seats.

Thats all by now  

Sweet! I will give you a real room (not the Revit included example) to look at soon. Would you be up to it?

I would also not add doors and shit, as you can easily fill those with customizable hook5 objects far better. I am just talking about the basic structure with textures. Furniture I would totally leave up to hook5. Maybe a central couch and other things, but not smaller, customizable objects.

[Sexvision]

Tried your room and converted without textures. As it is now it will use 18 fps without textures. thats actualy pritty ok. You will still have some room for textures.

Converting your room to h5m will take some time due to the poly count but still not that bad.

here is the file so you can test it yourself. Dont set alpha channel because it will not show anything.

untitled.rar

[HDiddy]

18 FPS seems high, very high for a room with no textures and under 300k poly, even for a Object Room.  I quickly spun up a test room and I was getting 71FPS with no textures using a standard VX Room.  The room is scaled pretty accurate on load so there was no need modifying scale.

I would say is very possible to use it, but you would need to account for the time to either group the 517 objects in the room or apply textures to each object.  Sometimes you may have export settings on the application that will group object nodes into something more manageable but it depends on what application you are using.  Attached is the room.

 

VX.TestRoom1.TestRoom.7z

[Sexvision]

56 minutes ago, HDiddy said:

18 FPS seems high, very high for a room with no textures and under 300k poly, even for a Object Room.  I quickly spun up a test room and I was getting 71FPS with no textures using a standard VX Room.  The room is scaled pretty accurate on load so there was no need modifying scale.

I would say is very possible to use it, but you would need to account for the time to either group the 517 objects in the room or apply textures to each object.  Sometimes you may have export settings on the application that will group object nodes into something more manageable but it depends on what application you are using.  Attached is the room.

 

VX.TestRoom1.TestRoom.7z 2.31 MB · 0 downloads

I meant i run empty at 120 fps took 18 of it got 102 left.

[Sexvision]

Did some texture work on your room. UVs are also ported so thats good. there all at 4k. that will take a little time to reset. i will test it to see what it will do on 4k.

[HDiddy]

9 minutes ago, Sexvision said:

I meant i run empty at 120 fps took 18 of it got 102 left.

Ahhh..ok that makes more sense.

 

3 minutes ago, Sexvision said:

Did some texture work on your room. UVs are also ported so thats good. there all at 4k. that will take a little time to reset. i will test it to see what it will do on 4k.

Nope no textures came over but a lot of the UVs overlap....so not sure how easy it would be to add decent textures without the original material to seperate all the overlapping UV's.

[Sexvision]

1 hour ago, HDiddy said:

Ahhh..ok that makes more sense.

 

Nope no textures came over but a lot of the UVs overlap....so not sure how easy it would be to add decent textures without the original material to seperate all the overlapping UV's.

I have used some tileable textures and im able to use alpha channel again after setting some textures. Boost me up to 120 fps again.

Ill going to sleep now. its way to late