It is hard to get models out of Blender into Godot (correctly and without loosing half the work).
Gamedev relevant open-source projects should come together to create a new 3D interchange format. More in the GitHub discussion.
I’ve submitted this proposal to other open source projects as well:
I feel like the problem is Blender and Godot have different needs when it come to models. Godot is real time and blender is slow but far more capable in what it produces. It would be extremely difficult to make a new format that can handle both use cases. I think what we really need is a “best modeling practices guide” for godot/gamedev. My biggest issue when I never know what will work and what doesn’t. I don’t think we need a new format just more documentation and guides for modeling in blender for gamedev.
I feel like the blender and Godot communities should have enough overlap to provide a steady stream of documentation and tutorials.
I would love to see a tutorial showing a minimal game built from the ground up including modeling, texturing, and animating models in Blender then transferring them to Godot and building the game. Instead people seem to focus on one app or the other and hand wave away the details of the transfer.
You can’t create 3d models in Godot. (To be pedantic you can, but it’s not made for that)
Blender can perfectly produce models intended for real time rendering. It also has the Eevee renderer^^ Whether a 3d file stores 10k or 10M vertices doesn’t matter. It’s on you, the asset creator, to know the requirements of your project/customer/employer. Blender, or any other tool, should not impose any artificial restriction on that.
The model made in Blender should arrive in Godot without having lost any data that both tools do support perfectly. The thing that holds the model while it goes from one to the other, either looses half the work, or Blender itself refuses to put in big parts of your work, because they weren’t able to fully and correctly reverse engineer FBX.
I’m not faulting Blender developers for the reverse engineering, I’m arguing for a format that does the basics right, isn’t proprietary, and hopefully extensible.
