Alpha channel is the part of the pixel data that determines transparency. You need to be using 32-bit color-depth to represent the alpha channel (8-bits for each of red, green, blue and 8-bits for transparency). The alpha channel is really a mask that specifies how the pixel's colors should be merged with another pixel when the two are overlaid, one on top of the other. In many graphics programs you create an alpha channel by first creating a mask and then saving it to the alpha channel. Black and white are off and on and any shade of grey determines the level of transparency in between.

Image formats like JPG are only 24-bit and don’t carry transparency information but formats like TGA are 32-bit and will carry alpha channel information. If you want to use alpha channel in videos, the video must be saved with 32-bit color-depth. Most are saved with only 24-bits. VirtualDub can do 32-bit saves. Also Ulead Cool 3D will create 32-bit video to use as overlays with an alpha channel. (I do this all the time)

As for the second half of your question, I'm not sure how the different settings affect it. When I use TGA files, I use the "straight (unmatted)" option and when I use video files I use "premultiplied". I don’t know why, but it just works that way. I haven't found a video codec that I can save alpha channel information with so I use uncompressed when I need 32-bit color-depth for overlays. Hope this helps.

~jr