I wonder if you could pack an alpha channel into say the blue or red channel? you could perhaps quantize the blue so it goes from 8 bits to 4 bits, then use the other four bits to make a mask and use the shader to unpack it. It should be quite simple to set up an effect in an NLE to ‘encode’ the channel this way. Or easier would be to make the blue channel a side by side dual image with the blue on the left and the alpha on the right - the blue channel would have a lower horizontal resolution but I reckon it would be fine. A similar quantisation happens when you record video anyway, like 4:2:2 sources - If you had a 4:4:4 codec then you wouldn’t even notice the difference.