If you're making title overlays for video, you'll need to know about alpha channels. They define what's visible and what's not within an image. Within the PSD format, alpha channels are fine, though when exporting, it can get messy.

Let's say we've got some text, with a layer style shadow, on a transparent background. To avoid mistakes and large files in the editing process, we want to use a common, flat format with an alpha channel, like PICT, from our PSD master. First, we need to find the alpha channel. Command/Ctrl-clicking the layer won't select the layer style, so we need to Create Layers to do this. Either duplicate the layer first or work on a copy. Now flatten those layers together, and Command/Ctrl-click the layer to get the transparency. In the Channels palette, click the New Channel from Selection button near the bottom.

Important: when flattening a layered image containing an alpha channel, Photoshop assumes that the background is white. This is called premultiplying the image (with white). But when we open such an image in a program that isn't expecting a premultiplied image (ie. most of them) we'll get a subtle white halo around the image. (Sometimes you see this on the web with transparent GIFs, though this is more subtle.) This occurs where the alpha channel has selected (say) 50% of the edge. This edge used to be 100% black, 50% transparent. After blending with white, it's become 50% black (grey) but it's still 50% transparent. Muddy grey. Yuk.

We have the correct alpha channel, though. What we need to do is change the image's transparency values from 50% black to 100% black. Since we know, through the alpha channel, what the transparency value should really be, the extra information in the layer itself is only going to confuse the issue later. Make sure the alpha channel is intact, then create 5 copies of the layer itself. Merge them all together. Create 5 copies of the newly merged layer, then merge them again. And one more time. The edges will look harsh and ugly, but the alpha channel will retrieve the correct values.

Basically, the transparency values in the image have been converted to 100% while the colours themselves have not changed. There's nothing important left for Photoshop to premultiply, and the colours will be correctly interpreted when we save a flattened PICT with alpha channels. This whole process can and has been actioned.

Note: there is a filter, part of the Filter Factory, to change transparency values to whatever you want, but unfortunately it isn't complete; when used in an action, it pops up its dialog every time. The current duplicate/merge technique is fully automatic. Since it merge-copies the current layer and pastes into a new image, it needs anchor points to ensure it lands in the right place. And what are they? On a background layer, the four corners of the image have 1% visible pixels applied with the pencil tool. Anchors. Very handy sometimes.