>>25500944
That's the ffmpeg encoder, usually apngasm is slightly better because the ffmpeg encoder just encodes each frame to PNG using ffmpeg's PNG encoder which sucks, while you can feed frames that you ran through a PNG optimizer to apngasm and it'll use them unchanged. And if you need to reduce color depth to 256 colors to reduce filesize, use pngquant which is better than other quantizers.
Most tools for manipulating APNGs were abandoned right after the format came out because it got 0 adoption, so although the format has potential, it has never been realized, and in practice it's worse than GIF.
Here's an animation I just made showcasing the different formats:
The smallest version (lossy GIF) of this animation was made by taking an anime clip, averaging all repeated frames and removing duplicates for better image quality, tiling all the frames into one image, giving the resulting image a single global 256-color palette with pngquant, taking out the tiles, and then running gifsicle on the result. The 2nd smallest (256 color APNG) had the same source as the lossy GIF but without the gifsicle pass, and it was assembled with ezgif (my god I hate SaaS) which produces smaller images than ffmpeg's APNG encoder. The 3rd smallest (lossless GIF) was made like the lossy GIF, but without the gifsicle pass. The 4th smallest (8bpc APNG) was made like the lossless GIF, but using its 16bpc source quantized down to 8bpc instead of down to 256 colors.
After all this optimization to reduce file size, the GIF version is still the smallest and looks good enough.
A webp version would beat all of those with the least effort though.
It's also worthy of note that the lossless GIF isn't all that much bigger than the 256 color APNG, looks exactly the same, and you won't have any compatibility issues like you will with APNG and webp.