Second post about input file formats, because I suppose this is the right place for any possible future improvements... 

This post is related to TIFF so called alpha channel, to multi layer PSD, to source images which include moving parts, to fisheye lenses and to multiplane projection...

Full-frame (180° diagonal) fisheye images are rectangular, but many other shapes occur between this rectangle and a circle, depending on the focal length and sensor size. PTgui have a tool for that (never used it.) When using PTA I prefer to leave the black parts in the source images and to remove them in the resulting multi layer file (can't use JPEG, must use PSD or TIFF.)
Pro: image center is not moved by approximative manual cropping of source images
Con: many black areas to be removed manually from final image, no Enblend/Smartblend use is possible

Its well known that cars, peoples, clouds, tree leaves, waves, birds and butterflies  tend not to be still when one is shooting panos. Using [b}output[/b] multi layered files for the resulting pano and one of the many PS-or-any-other tools is a possible solution. Being able to use alpha channel masked TIFF or multi layer PSD for input files is another possible solution.

What if I want to stitch several panoramas (this web page shows that this can be useful: http://www.panorama-numerique.com/squeeze/squeeze.htm , this Max Lyons forum thread Hybrid Rectilinear & Cylindrical projections http://www.tawbaware.com/forum2/viewtopic.php?t=2866 is about the same point and here is an academic approval: Squaring the Circle in Panoramas http://www.vision.caltech.edu/lihi/Demo … orama.html

(I suppose that the method which consists in prestitching several (cylindrical projection mode) columns and stitching the columns afterward to avoid large file size problems is useless when using Autopano Pro !)

In my opinion, a wide range of input files formats is something to consider: not having a high priority, but not to be forgotten...