Personally, I'd look for a kitchen funnel to act as a gravity feed hopper, then have a plastic tube which fits on the narrow part of the funnel/hopper which gets the beads into an individual line using gravity. Big enough in diameter so they dont block up but small enough sort them. It might also pay to have this bit vibrate to prevent any beads from jamming up as gravity takes them down the tube which could be flexible so that the end can then be married upto a switching mechanism which stops the beads from falling out.
This switching mechanism could be a couple of things, a simple rotating disc with holes in that one bead can fall through/into* at a time, or maybe a solenoid/piston which controls a flap that lets one bead out at a time. The latter I think could jam more often though.
*As the disc rotates underneath the disc could be another flat surface to stop the bead from the hopper falling through. As the disc rotates, you could have another mechanism which then does the sorting, moving tubes for different colours into position so that as the disc rotates the bead can fall through into one of the tubes moved into position for different bead colours.
The identification can be done by a camera looking at the bead as it rotates around the disc.
As ambient light ie daylight clouds passing in front of the sun on a summers day can affect the colour the camera would see, you could need to enclose the disc & camera into a box of sorts with a lamp inside to keep the lighting consistent. The for the software just position the camera so that it can see the bead, and then onscreen divide the camera feed into lots of small grids. Select the grids you want to watch ie where the bead comes into view, and then look for an avg colour for each of those grids. Set tolerances for a range of a each colour to allow for differences in dye strength when the bead is made and mixed. When a colour range is recognised, have some more software control the positioning of the tube the bead will fall into.
That should handle the overall operation of bead sorting, but variations of it could be used for anything really, sorting seeds or even small electrical components.
If you didnt have the rotating disc but a solenoid/piston moving a flap to allow one bead at a time out to fall through a transparent tube, you would need to be able to identify the colour in much the same way as before, ie divide camera input into grids, choose grids to watch and then id colour that way. Then you would need to have another device which could either stop the bead before moving a sorting tube into position, or make the tube long enough so that the machine can get the right sorting tube into position in time when considering the time it would take a bead to fall down a tube using the laws of gravity.