Loading Models Using Assimp // OpenGL Tutorial #18

[@nickpickering6620]

It took me over a week of rewatching this video to integrate the concepts into my own game engine, but finally got it all figured out. Thanks a ton!

[@OGLDEV]

Glad to hear, good luck with your engine!

[@1004-u5t]

nooooo please!! I have to do this in less than 12 hours my project is due:((((((((((((((

[@eliseoryez1196]

@@1004-u5t how did it go big bro?

[@michaeldpaul]

14:40 I wish more tutorials would tell you this! (and I wish I had found this video months ago when I was pulling my hair out). Thank you for all the great info you share with us

[@OGLDEV]

Glad it was helpful!

[@OGLDEV]

Some of the code in this video is going to be changed in the next few tutorials as we add material and lighting support. If you want to get the same version so that you can follow along with the video you can checkout the tag 'TUT_18_ASSIMP'.

[@holdipoldi6048]

this explanation is very helpful for me ! Thanks and greetings from Munich 🙂

[@OGLDEV]

You're welcome :-)

[@0xB16B00B]

im so glad can know about your channel, your video and how you explain something is very well and easy to understand. Thank you very much 🙏🙏

[@OGLDEV]

You're welcome :-)

[@MrFacciolone]

on some system enabling for DEPTH testing won't be enough, a depth buffer must be explicitely initialized when the OpenGL context is constructed. With freeglut it must be specified this way : glutInitDisplayMode(GLUT_RGBA | GLUT_DOUBLE | GLUT_DEPTH ); glutInitWindowSize(width, height); Adding GLUT_DEPTH to the code was necessary on my machine. Great tutorial series I am learning a lot thx for the good work!

[@OGLDEV]

Thanks a lot for your feedback. I added it on the entire code base and also updated the tutorial on the website with this info.

[@MrFacciolone]

@@OGLDEV my pleasure to be helpful! Thanks again for the excellent tutorial, I started this series 1 month ago with zero experience with anything 3D or GPU related, and I feel am learning very fast because of how well designed and thought out your tutorial is. Code is also very well written and complete and that's a significant bonus to the whole learning experience.

[@OGLDEV]

Thanks! Good luck!

[@OGLDEV]

The following comment got deleted by youtube for some reason: "I have a bit of a different setup where my model class stores multiple mesh objects. However when I load in textures through the model class they render as weird characters instead of colors but if I load them through the mesh object it works just fine. The way it works is I pass in texture objects to the mesh class to store and use. At first I thought it was a memory issue with the textures going out of scope when being passed to the mesh but it doesn't seem to be the case after using dynamic allocations. Any thoughts?". Seems to be some confusion with regard to the distinction between a class and an object. Loading textures through the class means using a static method which is not attached to any specific object. The most straightforward for this to work is for every object to load its own textures and bind them at runtime before the draw call. If you want to share the textures between the different objects (e.g. for memory optimization) then you can use a static map in the class and have each object access the textures that it needs using a common index. According to the bug description my guess is simply that the wrong textures were loaded. You can track the actual texture that is being used using apitrace. See my tutorial on debugging OpenGL for a few examples.

[@benceweisz1063]

From what I understand, your mesh class uses per vertex normals. If say I wanted to do per triangle normals and I had less vertices than triangles how would I approach the problem? Are most models used in industry designed to just use per vertex normals or is it common to see per primitive normals. Thanks!

[@OGLDEV]

Assimp provides the flags aiProcess_GenSmoothNormals (vertex normals) and aiProcess_GenNormals (face normals) that allow you to specify which type of normal to calculate. Note that if the original model was exported with normals then these will be used as-is. I guess the rational is to respect the way that the object was modelled. AFAIK most models today use vertex normals because they provide a more realistic lighting effect. But face normals can still be used if you want to achieve some kind of a low polygon or less modern feel.

[@toddwasson3355]

Very helpful, thanks. Got some models running in my D3D12 noob engine.

[@OGLDEV]

Great, thanks!

[@tanveerasif5978]

Hi, thanks for the explanation. I would like to ask a question, What advantage 'Array of structure(AOS)' bring over 'Structure of Array(SOA)' ? I understand the difference but didn't understand the benefit of one, over the others.

[@OGLDEV]

afaik, there is a performance advantage to SOA. It is easier for the compiler to optimize for SOA (on the cpu for sure but probably on the gpu as well). The vectorizing instructions are usually geared towards SOA. Try asking chatgpt: "Is structure of arrays better than an array of structures for performance on the gpu" ;-)

[@ПавелКиселёв-э6п]

Thanks for your great video. I have some trouble when using this code in Code Blocks IDE (Include\ogldev_math_3d.h|37|fatal error: assimp/vector3.h: No such file or directory) It can't find assimp/vector3.h, but it in assimp directory.

[@OGLDEV]

You need to add 'ogldev/Include/assimp5' to the include path in the IDE. It contains 'assimp/vector3.h'.

[@nicolassuarez2933]

The same company behind opengl and glft has almost no out of the fox support for gltt and glb??? I am really strugling here. Any help? Thanks!

[@OGLDEV]

Assimp does support both gltf and glb. Perhaps there's a problem with a specific model.

[@dengan699]

Oh wow thanks for this tutorial series. excellent job

[@OGLDEV]

You're very welcome!

[@bimDe2024]

we can also use vcpkg package manager for assimp

[@OGLDEV]

Thanks. I never tried that. Good to know that it's working correctly.

[@Julia-fp5zr]

Thanks for the video! I am still a little bit confused about the Basicmesh class and its functions. Where do I need to put the BasicMesh class? I can't find it in the "tutorial18" folder on Git. Is it part of the tutorial.cpp file? Or is it an own .cpp file? Or where does the whole Basic:Mesh Code you showed in your video go? Thanks for your help!

[@OGLDEV]

The link to the source code which you can find in every video description points to the specific directory of the tutorial. In addition, there are supporting classes in the root directory under 'Include' and 'Common'. The BasicMesh class declaration is in github.com/emeiri/ogldev/blob/master/Include/ogldev_basic_mesh.h and the implementation is in github.com/emeiri/ogldev/blob/master/Common/ogldev_basic_mesh.cpp. This class encapsulates the population of index and vertex buffers from the data loaded by Assimp, initializes the vertex layout accordingly and provides a simple 'Render' interface to be used from the main application. This allows you to load any model type supported by Assimp (obj, fbx, etc) and then render it with a single call.

[@yoonhakim2455]

Hello. Could you explain what 0 means in HasTextureCoords(0)? and how we are able to do mTextureCoords[0][i]? I looked at the documentation and it showed that mTextureCoords is a one dimension array.

[@OGLDEV]

The vertex may include zero or more (current maximum is 8) texture attributes. You need to call HasTextureCoords with the index of the coordinates and it will tell you whether it exists. I would have preferred an API that tells you how many tex coords exist but this is how it works. I'm only interested in the first set so I pass in zero. mTextureCoords is indeed one dimensional array but of pointers which means that every element in the array is the base of a different array. So the first index is zero for the first set of coords and the second index is 'I' for the coords of the current vertex.

[@yoonhakim2455]

@@OGLDEV Now I get it thank you. Just one more thing, do you still have the source code from the video? I looked at your github page but I couldn't find one.

[@OGLDEV]

The source code is here: github.com/emeiri/ogldev/tree/master/tutorial18_youtube. If you want to get the same version that was used in the video (the master branch contains many changes) checkout the tag TUT_18_ASSIMP.

[@yoonhakim2455]

@@OGLDEV Thank you. your videos are awesome.

[@OGLDEV]

Thanks!

[@unluck7]

how do we import assimp 5.x or higher in unity?

[@OGLDEV]

Assimp is a library for loading many file formats. You can easily integrate it into your own application. Unity also supports several formats though I doubt that they are using Assimp. They probably implemented their own loader.

[@nicolassuarez2933]

Outstanding! I cloned your repo, so in theory I dont have to do anything in order to run tutorial 18? The first tutorials are working perfectly. But 18 is not working for me in version 2022. Also where the is the spider.obj? Thanks! 1>C:\Users\camis\Downloads\ogldev\Common\ogldev_basic_mesh.cpp(22,10): error C1083: Cannot open include file: '3rdparty/meshoptimizer/src/meshoptimizer.h': No such file or directory

[@OGLDEV]

1. The repo should be working out of the box, including tutorial 18. 2. For the missing meshoptimizer files run the following in the root dir: git submodule update --init 3. The spider.obj file is in ogldev/Content.

[@nicolassuarez2933]

@@OGLDEV It worked! I created a folder in ogldev\Include\3rdparty\meshoptimizer\src\meshoptimizer.h from git hub. How to load custom models from blender? With standard export parameters obj is not working for me. fbx, gltf, custom textures? Thanks!

[@OGLDEV]

With obj make sure to also 'triangulate faces'. fbx requires 'apply scaling=fbx custom scale'. Also triangulate faces. Not sure exactly about gltf but probably along the lines of the above. These things something need trial and error. obj is the simples to debug because it's a text format so you can manually tweak it.

[@nerijusvilcinskas7851]

Great video! Although, I have used interleaved buffer (Array of Structures) in my model loader using assimp and it works perfectly, plus I made it a template class and I can use different vertex attribute layouts, so would I benefit if I used SOA (Structure of Arrays) assimp loading instead of AOS?

[@OGLDEV]

There are various resources online that discuss the performance implications of AOS and SOA but personally I'm not aware of a benefit of one over the other.

[@nerijusvilcinskas7851]

@@OGLDEV I guess I'll stick to AOS for now because it works very well too, thanks!

[@deathspreads]

I'm trying to use this exact tutorial as the base code for a personal project of mine. Is there a separate repo for this project only? Otherwise I've been messing around with the code for an hour or so now and I just can't seem to make it run; there's only an empty shell window that pops up for half a second and disappears.

[@OGLDEV]

I prefer keeping everything in a single repo because it reduces the maintenance effort. Sound like you are on Windows so you can open the Visual Studio solution at 'ogldev\Windows\ogldev_vs_2022\ogldev_vs_2022.sln' and go to the 'Tutorial18' project. You will find all the cpp files that you need to build there. Note that some of them are actually located in 'ogldev\tutorial18_youtube' and the rest are in 'ogldev\Include' and 'ogldev\Common'.

[@deathspreads]

@@OGLDEV Yes I have managed to find the solution and I built and ran it, but I feel like I'm misunderstanding something, since all I see is a black console screen. I have watched the video but I don't seem to get how to make it run properly. I see that the arguments fed into the executable call are used in the glutInit function; does this mean I need to feed it arguments for it to load a certain model? It's either that or I mismanged my installation of the dependencies such as GL and ASSIMP.

[@OGLDEV]

The arguments to glutInit are used just for the initialization of GLUT which handles windows creation, keyboard control, etc. The actual loading of the model is in tutorial18_youtube\tutorial18.cpp:108. If you are not familiar with GLUT then I suggest going through the tutorials one by one. I've just verified that the latest code works ok so if you are not seeing the spider then this will require some debugging on your end.

[@greengem8537]

How can i see original code that was in video with tag?

[@OGLDEV]

git checkout For example, to get the code for this video: git checkout TUT_18_ASSIMP Most of my videos have tags and you can usually find them in the video description. I'm still working on adding tags to the rest of them...

[@dev.ashesh]

Will this work if lets say a custom binary file that contains a model inside?

[@OGLDEV]

What do you mean by 'custom'? Assimp supports many binary file types such as fbx. If you can load a fbx file in Unity there's a good chance that Assimp will be able to load it as well.

[@dev.ashesh]

@@OGLDEV like custom file format.. its a binary file that contains a 3d mesh inside.. but the problem is i have to find the values manually in hex.. and its very time consuming..

[@OGLDEV]

Then afaik the answer is no. It doesn't provide a general method for specifying a custom file format like a template or something like that.

[@danielkolev7458]

I have a question, how can I add my own model to insert in the code. Please I am just new and don't know the subject very well. Thanks in advance!

[@OGLDEV]

From this tutorial forward all models are loaded by creating an instance of the BasicMesh class and calling the function LoadMesh("/path/to/model/filename") on this instance/object. In this tutorial you can see the example in tutorial18.cpp lines 106-108.

[@BlueProgamer212]

I have assimp-5.2.5\build\bin\Debug\assimp-vc143-mtd.dll but not assimp-vc143-mtd.lib. Should I downgrade? I don't want to load the DLL files during runtime to access the ASSIMP functions because I noticed it takes a while to load. I'm building it with Visual Studio 2022 is this fine? I'll attempt putting it the same folder as my executable. I apologize if this is a stupid question. Thanks in advance for being kind enough to answer! EDIT: Found the answer that works for me, for anyone having the same issue, here's the solution: Just disable shared libraries to use .lib instead of explicitly loading .dll for the symbols Here ya go: cmake assimp-5.2.5 -G "Visual Studio 17 2022" -DBUILD_SHARED_LIBS=OFF

[@OGLDEV]

assimp-vc143-mtd.lib is created in assimp-5.2.5\build\lib\Debug. You can use the DLL by simply linking your app against this lib file and you don't need to explicitly load the symbols from the DLL.

[@gujeg]

Thx, you helped me really nice 🙂

[@OGLDEV]

Glad to hear that!

[@kimkulling]

Great tutorial, thanks for the great work!

[@OGLDEV]

Thank you for Assimp!

[@grengren6421]

Thanks for another great video.

[@OGLDEV]

You're welcome :-)

[@mosesmodise4998]

what happened to the shaders?

[@OGLDEV]

This video focuses on loading the vertices and their attributes using Assimp. Shaders were covered in tutorial #4 and from that point in the series we've been busy improving the shaders to include the relevant transformations as well as texture mapping. All this was done using some hard coded vertices and indices. So what we do in this video is replace these vertex/index arrays with vertex data loaded from model files. So basically there is no change in the shaders compared to the previous video (you can find them here:github.com/emeiri/ogldev/tree/master/tutorial18_youtube). In the next video we improve the shaders with initial material and lighting support.

[@bestapps4709]

how to load in Opengl es android java?

[@OGLDEV]

I haven't tried anything except C++ but there are java bindings of assimp: github.com/assimp/assimp/tree/master/port/jassimp. The Android build instructions are here: github.com/assimp/assimp/blob/master/port/AndroidJNI/README.md. Hopefully you can find a working example online.

[@bestapps4709]

@@OGLDEV thanks for answer buddy, I use my cellphone to create apps in java and Opengl es, and I need a code or a lib to put directly in the paste lib to begins to use it. I use AIDE and Sketchware apps to programming.

[@OGLDEV]

I would love to expand the channel to the Android space at some point but right now I just don't have the bandwidth. BTW, OpenGL vs OpenGL ES is not really relevant at this point. You need to be able to access Assimp and load your models from Java. Once you get this working you can use whatever API to render. ES is a native citizen in Android so I guess it should work.

[@bestapps4709]

@@OGLDEV Ok buddy I will try, I need really load my objects in ES. Have a nice day.

[@Byynx]

Thanks for your work but of all the tuturials you had to pick this one to show the SOA. It makes more confuse to understand assimp.

[@OGLDEV]

Sorry, that's the implementation that I currently have. But understanding Assimp is more about the data structures of the library. I'm storing the vertex attributes in different vectors that are then loaded into dedicated vertex buffers but you can pack them together in a single buffer and set the vertex layout accordingly.

[@UnrealCatDev]

9,11k subscribers? niceee

[@OGLDEV]

wow...

[@MrSalireza]

Awesome job ❤

[@OGLDEV]

Thanks!

[@joseplano4789]

Please do one loading animations on assimp

[@OGLDEV]

I plan to do a tutorial on skeletal animations but there are a few topics (e.g. lighting) that I feel I must cover prior to that. In the meantime, you can check my skeletal animations tutorial on the website: ogldev.org/www/tutorial38/tutorial38.html.

[@OGLDEV]

The skeletal animation tutorial is now available: kzbin.info/aero/PLA0dXqQjCx0TKQiXRyQU62KQgcTE6E92f

[@joseplano4789]

@@OGLDEV thank you very much for this

[@matthewexline6589]

Hey guys! "ai" as in "aiscene" stands for "ass imp" not "artificial intelligence"!!! Can I get some likes over here for figuring out that bit of pointless trivia for myself or WHAT?!?!

[@OGLDEV]

"artificial intelligence" is a clickbait...

[@shivanshuraj7175]

how your build assimp was so fast..what hardware you are using...i am doing it on 1155g7 and it takes 30-40 sec

[@OGLDEV]

It's just an Intel i5, but I did a fast-forward during editing to save your time ;-)

[@GBSCronoo]

Thank you!

[@OGLDEV]

You're welcome!

[@VerzatileDev]

I'm not sure why it was all skimmed through with the tutorial. Isn't it important to know how to add files to your projects ? instead of pretending that everyone knows where and how to put them. ( If this is some tutorial series where you start from episode 1 to which ever this one here is, then okey I understand) " Though if you are looking from the side you really don't have an idea what to do, well I will figure it out, but I find that a bit odd from this video " though thanks anyhow.

[@OGLDEV]

The target audience for this series is people who want to learn graphics programming with OpenGL and have a basic experience with C++. I'm using very basic features of C++ so you definitely don't need to be an expert but some experience in developing code and creating a project in your preferred dev environment is required. If you feel you need some ramp up here then I highly recommend watching one of the several C++ courses at kzbin.info. The courses for beginners cover everything you need to know in terms of creating a project, adding source files, building, etc. In the first few tutorials on this series I discuss some issues that are relevant to your question. At the end of "Creating a Window" (kzbin.info/www/bejne/eorYfXuIqNuNrc0) I talk about how I build stuff on Linux. If you are on Windows you may want to watch kzbin.info/www/bejne/jX63qYxnacZ9gKc where I explain how to create a Visual Studio solution for OpenGL programming. Again, this is not a full tutorial on VS so I just explain the specifics of getting stuff built and linked for OpenGL.

[@werenter]

Why you are using one of the worst linux distros? ArchLinux, Debian and Gentoo much better.

[@OGLDEV]

I'm used to it and it's been working ok for me. May try the others if I ever find the time.

[@kennedywee]

"Can't load texture from './wal67ar_small.jpg' - Unable to open file" Hi, thanks for you wonderful tutorial! I cant figure out why this happened since I'm new to programming. I wonder what I did wrong?

[@OGLDEV]

This file is located in the 'Content' directory. Which model did you try to load?

[@kennedywee]

@@OGLDEV "spider.obj" and I tried to load the other model inside Content directory but same error with the texture.

[@OGLDEV]

Are you on Windows or Linux?

[@kennedywee]

@@OGLDEV windows 11😄

[@OGLDEV]

It was a bug in the path handling on Windows. I pushed a fix. Thanks for the feedback!