A headless blender that can be run in the browser with webassemlby would enable services that use blender for calculating complex 3d tasks by running them on the client system. I can imagine many scenarios where this can be a huge game changer. For example rendering images that have been customized by the client with only some simple parameters.
I have started to look into how building headless blender to webassembly might be done, but quickly realized that this is far out of my capabilities. Thats why im reaching out to you fellow developers.
First of all I would be interessted if anyone can say if this is even realistically possible without rewriting blender.
Also, do you have some valuable resources that might help me in achieving this? Like a explanation of the build scripts/build files. Or Compiling C++ to WebAssembly.
Is anyone already working on something similar?
I hope this is the right place to ask. And that my interest is understandable.
I looked at this last year, got it to the point where a basic python script could run, reasons why i never continued the work are
The GFX stack in blender is highly reliant on opengl 3.x , the browser offers only opengl ES which blender currently does not support.
Even with a lite build the time the linker required to generate the code was through the roof (minutes per try,vs seconds for native code) to the point where no productive development could be done
doesn’t really matter all that much as this point, all drawing code in blender is targeting desktop openGL, reworking that to support any other gpu stack (doesn’t really matter if that stack is OpenGL ES or WebGPU) is a monumental amount of work.
Since it’s been a couple of years, I dare ask. Theoretically would the Vulkan port allow for this? As I understand it, Vulkan is built for that sort of gpu-stack-agnostic system, right?
