I’m back on my release early and often schedule… or at least something resembling it. This is a big bug fix release and officially brings the core of VapourSynth to beta quality. Less memory will be leaked and pressing F2 in VirtualDub is no longer the fastest way to run out of address space and crash.
There are only a few more things to add and change but the goal is getting closer. Report any bug you find no matter how small it is. The focus of the next release will be on performance and I’m interested in all speed comparisons between VapourSynth and Avisynth, especially the single threaded versions.
You should also check out Chikuzen’s GenericFilters which implement almost all missing features from masktools that aren’t already present in the core. This has made it possible to port a few more scripts to be fully native.
If you’re curious about the future direction and changes go check out the bug tracker. I use it as a place to write down most planned enhancements as well and scribble a row or two about it. I have also been commissioned to make an image writer plugin based on ImageMagick so that will most likely show up before R19.
Is there a reason to use ImageMagick, vs something like GraphicsMagick (http://www.graphicsmagick.org/)?
I haven’t tested it myself, but I believe Flickr switched over to it due to a significant performance improvement over ImageMagick.
If I remember correctly, the GraphicsMagick author does not consider every ImageMagick feature important, so not all features are implemented. I don’t know which ones and how useful they are, though.
I think imagemagick supports the most formats for reading and writing of the two, that’s the main reason. I may be able to make it work with both since they’ve got similar apis as well. Personally I find the handling of different bitdepths in both libraries weird and disturbing. Why do I have to decide the maximum input/output bitdepth at compile time? Especially when graphicsmagick claims to do everything internally with float precision if I’m not mistaken.
Correct me if I’m wrong about this detail.
On the IM download page, there’s a claim that the Q16 builds will use twice as many resources as the Q8 builds. As I understand, if you don’t need 16-bit precision, you’re better off with the Q8 build.
That’s the reason given for it to work like that, at least.
I do need it. Higher bitdepths and all that.