Hmm. That sounds like maintaining the alignment of the three beams would be a total nightmare.
it's funny, we're building this set up (https://github.com/PRNicovich/NicoLase) in our lab for microscopy. From the docs:
"When aligning your laser launch, do not rush. This is not something to attempt late on a Friday afternoon in a free hour before heading to the pub. This is a task involving a great amount of precision and small, careful movements. There will be very little in the way of positive results until the task is nearly complete. Set aside up to a several hours to complete the alignment, especially on your first try"
In lieu of a cube, you could use three (R,G,B) dichroic notch reflectors (mirrors) (2) and combine the lasers 'manually', but the cubes are nice because you don't have to build a mount for the I individual mirrors.
You can either scavenge a cube or dichroic mirrors from an old standard projector, or pick them up online for a few bucks. After that, just place the combined beam in the same path at the red laser shown in this example set-up, modulate the three lasers according to an image's RGB intensity values (you can separate an image frame into three RGB frames using openCV), and there you go.
Tangentially, the basic idea of scanning a laser to form an image is what underpins the Virtual Retinal Display (VRD) - a head up display technology developed at the Univ. of Washington and later attempted to be commercialized by Magic Leap (ML) for Augmented Reality eyewear. Even though ML hasn't panned out, I still see the VRD as a very viable path forward for high resolution/field of view AR displays.
1. https://en.m.wikipedia.org/wiki/Dichroic_prism
2. https://en.m.wikipedia.org/wiki/Dichroic_filter