@Muffinshire
Looks like it got did. Thanks for the tips, and seconded: it’s especially important if you’re dealing with pre-G3 stuff from ropey filmstock sources.
I think dithering method is perhaps more important than you make out. At least in GIMP, the Positioned or None methods are gonna work better than the Floyd-Steinberg alternatives because the F-S ones are highly context-sensitive and own-output sensitive. Any small noise can really throw them out across a wide area. Simpler schemes will be predictably locally the same from frame to frame.
Another trick is to save as GIF, optimize it with gifsicle or some other tool, then load it back in and erase the noise that’s just become visible with an opaque brush or selection. Aim for large areas of the transparency colour, and erase anything that’s not motion or backfill.
It’s the video noise that really hikes the file size. GIF has a very crude form of compression that records every pixel that changes between frames, so noise effectively makes a large proportion of the pixels change from frame to frame, bloating the file size. Running a despeckle filter can help, as can choosing the dither mode carefully, but the real savings come if you can do a bit of manual tweaking.
What I do is freeze areas of the image that don’t change with an overlaid layer, or use an underlaid layer and only retain the changes from that “base” on subsequent frames, which greatly reduces the inter-frame noise. It’s a bit time-consuming for complex scenes (and not at all practical if there’s camera movement), but the results are much smaller. Give me a few minutes and I’ll work some voodoo on it.
Colors are optimized. Fewer colors proved to reduce quality severely, without noticeably reducing file size. I already manually removed all duplicate frames and reset the frame rates accordingly.
The only thing I’d suggest is resizing it. Good luck!
Looks like it got did. Thanks for the tips, and seconded: it’s especially important if you’re dealing with pre-G3 stuff from ropey filmstock sources.
I think dithering method is perhaps more important than you make out. At least in GIMP, the Positioned or None methods are gonna work better than the Floyd-Steinberg alternatives because the F-S ones are highly context-sensitive and own-output sensitive. Any small noise can really throw them out across a wide area. Simpler schemes will be predictably locally the same from frame to frame.
Another trick is to save as GIF, optimize it with gifsicle or some other tool, then load it back in and erase the noise that’s just become visible with an opaque brush or selection. Aim for large areas of the transparency colour, and erase anything that’s not motion or backfill.
Feel free to reupload it to DB - I’m not staking any claim on it here.
What I do is freeze areas of the image that don’t change with an overlaid layer, or use an underlaid layer and only retain the changes from that “base” on subsequent frames, which greatly reduces the inter-frame noise. It’s a bit time-consuming for complex scenes (and not at all practical if there’s camera movement), but the results are much smaller. Give me a few minutes and I’ll work some voodoo on it.
Colors are optimized. Fewer colors proved to reduce quality severely, without noticeably reducing file size. I already manually removed all duplicate frames and reset the frame rates accordingly.
The only thing I’d suggest is resizing it. Good luck!
Likewise. Assuming I can shrink the filesize down to something reasonable.