The conventional wisdom says that gentle gradients of color don't render well as gifs. Sometimes it maybe true, depending on the rest of the image, but there are WAYS. I've made delightful gradients in a GIF with only 32 colors or even 16 if dithered.

One key I found to getting a good GIF is making a selection on the tricky areas of the image while it's still in RGB. If you leave a selection active while you convert to GIF, Photoshop's algorithm will weight the histogram toward the pixels in the selection when it calculates the GIF color table.

This method can often eliminate visible banding, nasty color shifts and other ills. It's been a little known feature since Photoshop 3.

Another trick in RGB is to select all the dark tones (the ones that look black but might even be 60-48-56) and replace them with one color.

Then do the same for all the light tones. This frees up many cells in the GIF color table for colors where the eye can easily notice the difference.

Sometimes dithering just one part of the image is the key. I do a cut and paste of the part of the picture I want to dither, and index just that part as a new untitled image. The remainder part I index as well without dithering, and then take it back into RGB.

Now when I paste the two parts back together, I already know the maximum number of colors I need, and the odds are it can be just a bit lower.

With these methods I've been able to chop 70% off a jpeg file size, and get a sharper picture as well. And since they are GIFs, I also have the option for transparency, animation, etc.

One online example is at www.peacesite.org. There is not a jpg on the page, and every graphic has gradients.

The trick is to take the art work into Lab color space. In Lab, almost all the detail is in the L channel, so you can go to the a and b channels and run a very high Gaussian blur
on each one -- sometimes over 3 pixels. The detail is still held in the L channel. You can even apply some unsharp mask or contrast to the L channel to jazz up the image as needed.

Some images have sharply contrasting colors next to each other and the blur on the a and b channels becomes visible at the edge between them. Just select six pixels or so around those edges and invert, so the blur doesn't get applied to the edge.

When you return to RGB, the JPG algorithm can now crunch the file a lot further before you see any nasties. I just finished preparing images from a professional photographer for the web. I had to keep the file sizes reasonable AND keep his work looking good. These techniques plus the plug-in worked miracles.

By the way, JPG plug-ins make tricky little images -- they're like tightly compressed springs. Just doing a SAVE AS and renaming the file will often cause the file size to double or more, and you haven't touched a single pixel!

The only way I've found (in Windows) to rename and still retain the small file size is to rename directly in the folder using Windows Explorer.

Hope thes tips are helpful.