On May 24, 2007, at 9:06 AM, Paul Miller wrote: Micah Sharp wrote: Hi, what are the possible differences between the thumbnail render and the canvas render? I am getting wildly different results between the two when adding my plug-in on top of a 16 bit per channel tif file that is text. An 8 bit per channel tif file of text has no difference or problems, nor does applying my plugin to Motion's built in Text layers. Any ideas? The same code executes without any problems in FCP.
Looks to me like the thumbnail is being rendered in 8 bit, and the canvas is rendering in float? (I didn't think there was a 16 bit rendering format in there, right?). Looks like massive wrap-around in the canvas. Does your plugin support floating-point software rendering?
This is correct. The small preview is always 8 bit. If your filter is software only, then you will only ever get 32-bit per channel floating point data, even if the project is set to 16-bits per channel. The reason is that to do anything with 16-bit floats, you have to convert them to 32-bit floats anyway, so we just do it for you.
However, if your plugin is running a fragment program, then the fragment program will get 16-bit floating point data in this case. Of course, you shouldn't have to do anything special to your fragment program to handle the 16-bit case. It should just work. (Well, with the exception of bilinear interpolation which works on some cards at 16-bit and not others, but this doesn't look like that issue.)
Darrin
|