VP9 thread, this time with bloated sample MP4s at 720p because H264 blows hard at 1080p and above.Windows batch script found in webm related to use as a reference point:ffmpeg -i in.mp4 ^-c:v libvpx-vp9 ^-b:v 0 -crf 30 -g 300 ^-cpu-used 4 -lag-in-frames 25 -aq-mode 2 ^-tile-columns 1 -row-mt 1 -enable-tpl 1 ^-an out.webmREM -profile:v 2 -pix_fmt yuv420p10le ^
First 20 second 720p sample from the anime: After the rain.
>>107119169Output VP9 transcode using OP params.
>>107119169>>107119179FFmetrics of these 2. Not sure what scores to aim for but they already seem pretty fucking high from the get-go.
I'll focus on VMAF since that seems to be the most trusted in the video autism world. I found this:>Our premium quality setting, boasting a VMAF score of 96.45% and an encoded size of 207.66 MB, provides unparalleled visual fidelity suitable for high-end applications where pristine quality is paramount. This setting achieves a compression size of approximately 87%, demonstrating a substantial reduction in file size compared to the original 1.6 GB.>The optimal quality setting, with a VMAF score of 93.06% and an encoded size of 85 MB, strikes an excellent balance between quality and file size reduction. This setting achieves a compression size of approximately 94.7%, catering to a wide range of viewing environments and devices.>Lastly, the standard quality setting, with a VMAF score of 90.07% and an encoded size of 56 MB, offers a compelling option for scenarios where conserving bandwidth and storage space is essential, without compromising on perceptible visual quality. This setting achieves a compression size of approximately 96.5%.https://www.slashed.cloud/blog/slashedcloud-vmafTL;DR seems to be:>VMAF scores above 95 indicate very high quality but result in ballooning file sizes>VMAF scores below 90 are YIFY-tier>VMAF scores between these (ie 93) strikes a compression/quality balanceSo I'll modify the CRF parameter until I get VMAF 93 for the first video sample.
I think that animu or cartoon is possibly the worst kind of material for such comparisons, you only see how the edges are affected, while the textures are flat so you can't tell anything about them. Some irl shit would be much better, forest or any other heavy foliage.I've been mostly rendering fast game stuff and for size limited situations I find x265 superior to anything else. It has the best balance between blur and pixelation. x264 is way too pixelated and it ruins the edges the most, but at least it renders fast. VP9 and AV1 in limited size scenarios provide insane amount of blur, they are probably only good for porn, since blurred skin sometimes even looks better than the sharp one, also they render very slowly.For 4chan I'd use VP9 for static cartoonish stuff and porn, while x264 would be better suited for some heavy movement irl stuff or gameplay where preserving the sharpness and depth is more important.
>>107119654VMAF is supposed to be agnostic to all that but practically speaking 4chan's tiny 4MB file size limited kinda does void motion intense footage uploads here. Anime is great for testing video codecs in general since they can flex their bigger macroblocks there. For example H264 has a macroblock limit of 16x16 while VP9 can go all the way up to 64x64. However outside of things like Anime video bigger macroblocks aren't as useful.
CRF 40 = 90.7 VMAF @ 400 kbps & 0.9 MB file sizeCRF 38 = 91.6 VMAF @ 440 kbps & 1.0 MB file sizeCRF 36 = 92.6 VMAF @ 500 kbps & 1.2 MB file sizeCRF 34 = 93.7 VMAF @ 530 kbps & 1.3 MB file sizeCRF 35 = 93.1 VMAF @ 510 kbps & 1.2 MB file sizeWebm related is last one to keep things consistent.
>>107119706Testing on anime will always make the VP9 look better, because of the flat textures. I trust my eyes more than any artificial method of comparison.
>>107119866Not this time. Something is VERY wrong with the OP settings.ffmpeg -i in.mp4 -c:v libx264 -crf 28 -preset medium -an h264-crf28.mp4CRF 28 = 93.4 VMAF @ 350 kbps & 0.9 MB file size
Welp, seems like this isn't going to be as straightforward as I thought. I'm calling it quits for now, if this thread is still up tomorrow I'll continue finding out what the CORRECT VP9 parameters are.
>>107119948Can you test with more than one example?
>>107119983Ye but I'm boofed. I expected OP params to be enough for a like 20-30% better compression efficiency vs H264 as normally expected but something is wrong.VP9 is already like slow as balls, I hope I don't have to resort to 2 pass...
Double file size and no audio hail the eternal king H.264!
I just open Shutter Encoder, convert to AV1 and done
>>107119158Hooly shit windows is miserable. I forgot about the "New Text Document" and save as lmao. Could have opened Notepad but I dont blame you at all, Windows UX sucks.Also dont add C:\ to your path, you could have created some directory at least you heathen.
this retarded fucking thread again
>>107119158>set environment>whole C: drivefucking hell, op. do better, faggot.
