REALLY high quality videos. Upscaled is fine as well.Preferably anything that was 4k60FPS before it was sized down to fit /wsg/.This board has a 6MB limit (with 2048X2048 pixel limit) (hopefully the retarded admins can actually update that someday), so 15 seconds is probably the upper limit to prevent too much downscaling or downsizing.If you want an easy script to downsize things for /wsg/, I wrote this script with ChatGPT. Instructions are in the .bat file, but just put in a file path or drag-and-drop a file and hit enter and it will iteratively downsize it for /wsg/. There are other flags for other boards (4MB, no audio, etc.). This just reduces the dimensions to keep it simple.Script - https://files.catbox.moe/g9d7bb.rarLet's got for visual quality rather than length. Keep everything short.
Old anime doesn't always upscale well
Hopefully this modern anime looks a bit better
>video >60fpsGamer scum should kill themselves
>>6244841Every thread has at least one person who only complains (usually about something retarded) and never contributes.
>>6246595Bleh, had an extra frame in the beginning.Fixed.
>>6244841Hey, don't blame gamers. I play tons of games and know not to fuck with the framerate of animation. Blame retards.
I'm here trying out my own encoding method
>>6244233I highly recommend android phone posters use WebMChan
>>6244303Pretty neat!I saw a webm of the android saga that looks amazing awhile back. It looked like a flashback scene.
>>6256726That was probably me, I doubt anyone else is autistic enough to upscale DBZ clips and then downscale them again to post on 4chan. I've upscaled a lot of anime stuff, and AI has evolved so fast that stuff I upscaled 3 years ago looks like shit compared to something now.Even now, it is rare that it does a really good job.
AI is good at slow shots. Increasing FPS with fast motion can get very dicey. Eventually it will get better at that, obviously.
>>6257205
>>6257205Ironically not a good test, the text on the train had some motion blur near the end.
>>6257059The only place I want more frames is in panning shots. Pretty much everything else can fuck right off.
I don't know if you'd call it high quality, but I was bored and wanted to test encoder limitsUHD Blu-ray REMUX source1/2 resolution scaleHDR>SDRlibvpx-vp9 2 pass at 3.05mbpsnlmeans s=4.0:p=7:r=5
>>6260572>>6260573Nice, I have GBA
Let's see how far we can push an encoder with a fairly clean, digital, noise free source. Kind of a shame that what we have to work with is VP9 and a 6MiB size limit. If only we had HEVC support or at the very least a large file size limit. Truth is, you're just going to have overly aggressive quantization at these bitrates and there's not a damn thing you can do about it. Oh well, you do what you can!
>>6260632If only Hiro actually gave a shit and updated the fucking code.6MB is ridiculous, literally a SINGLE TEXT ON A PHONE can have more than a 6MB clip now.
>>6260688Yeah, it's the file size really. I can understand them not adding H.265 support due to the license clusterfuck and the compression efficiency of VP9 is already close enough, but 6MB is fucking ridiculous. This is 1.9mbps and I guess it's acceptable, but you gotta use libvpx-vp9 and do 2 passes and it's so painfully slow.
>>6260579Nice. I need to make more console intro webms. I made one at 4k, but apparently 'image resolution is too large'. I wish Hiro would at least update the boards.json to actually reflect the true file parameters.
Let's see how far x264's default CRF of 23 can be pushed while using the veryslow preset
>>6266150damn, shrank way too much
Just infodumping from my own experiences>If you keep clips as short as possible, you can get away with CRF encoding and use a relatively low value like < 18>downscaling to 720p is worth it because you'll get big space savings at the same CRF>depending on your source, play around with -tune film or animation (grain is usually too expensive but can be worth it) >use the slowest preset you can stand
>>6267940You're barking up the wrong tree with CRF.https://chameleon-ai.github.io/webm-guide/#crf-vs-average-bitrateAnd for 6MB the math works out to about 12 seconds for your average 1080p video.
>>6268322>6MB, 12 seconds, 1080pThat's like what, 4mbps? Not enough for *high quality* 1080p imo. Even with libvpx-vp9 and I avoid that like the plauge because my god is it slow, especially if you do 2 passes. I said you could get away with CRF encoding if your clips are short and if you downscale, not that it's optimal because if you're at 720p and your clip is ~8 sec, you can probably just wing it with x264 at CRF 17. But yeah, I will agree that you're more often that not better off just doing 2 pass with a specified bitrate.
>>6268417Depends on what you mean by high quality, that's roughly the video bitrate of 1080p youtube videos. If you don't consider that high quality, fair enough, but for those encoding from youtube sources, anything higher is placebo.A 4k bluray is like 50,000 and if you consider that truly high quality, that's 1 second not even including audio.>>6267940 is just under 10,000 and >>6244233 is 3,311 and it still has headroom to get to 6MB
>>6268621Oops, I didn't specify. 50000, 10000, and 3311 kbps
>>6268621Yeah, the OP video appears to be x264 with the veryfast preset with everything else default based on the metadata. I just use veryslow for most videos I encode for here because1. the compression efficiency really helps at these low CRFs/mean bitrates2. How much slower is encoding a 15 second video anyway?I get what you're saying about yt as a source. I encode these from remuxes of blurays, so I've already got a nice 30+mbps H.264 source to work with and imo playing around with presets, codecs, tunes etc. can be worth it. Honestly though, x264's defaults are pretty good and it's hard to make it look bad. Still, I stand behind downscaling to 720p because it allows me to drop the CRF. For 1080p, the CRF/bitrate goes up too high and the quantization jumps up too much and I'd rather take a slightly blurry 720p video than a bit starved 1080p one.
>>6269012meant to say high crf/low mean bitrate a few times. sorry, i are is retard.
>>6244233
>>6267940>>6268322>>6268417>>6268621I hope it's okay if I ask you guys, since you seem to know a bit about video encoding and such. But is there any good lightweight programs out there for making webms?A while back I tried something called webm for dummies. But I felt like it didn't do a very good job, at least not compared to some other webms I see other people upload. I also tried encoding a few clips with scripts using ffmpeg. But that turned out even worse, with lots of artifacts and the filesize didn't even stay below 4mb.Just curious if there's some new and simple program out there that can help you make webms, as I'm not very good with using scripts and calculating bitrates / length / resolutions etc.
>>6269685If you're GUI oriented, WebmGenerator is a pretty popular option: https://github.com/dfaker/WebmGeneratorI'm the developer of webm-for-4chan and I don't even think about making webms anymore, it's now trivial for me. If you're not averse to the command-line: https://github.com/chameleon-ai/webm-for-4chanFrom what the WebmChan for Android developer has posted in various threads, I think it produces fairly good quality webms but it's still in early development: https://github.com/jpeg91/WebMChan/releases/Other options like handbrake and rawdogging ffmpeg are more for power users who know what they're doing, since all the calculations are manual.
>>6269700I wasn't expecting to get an answer so quick. Many thanks for your reply, as well as the links. I'm definitely going to take a closer look, and I truly am grateful for this anon, thank you!>handbrakeAh yes. I remember trying that one too, and it did do a pretty good job. It's just that I'm terrible when it comes to using command-line, scripts and doing it manually. I honestly can't do any movie / video editing at all, although I've been wanting to learn when I was younger. But I'm way too green at this, as well as dependent on using some type of GUI as support-wheels, at least for now.I do have a question regarding converting videos into webms, and perhaps you or someone else here can shine a light on this.What I would like to know is. Does the source video file have any effects on the output file when encoded to a webm or mp4?What I mean by this is. Let's say, for example, that I record gameplay footage from a game in 1080p and 60fps, but maybe I want the webm to be 720p and 30fps to help lower the filesize. Is it okay to record the footage in the higher resolution and frame rate, before converting it down or is it better to have the source recording already be in the desired 720p and 30fps first, before converting it into a webm, or maybe it doesn't matter at all?Yes, I am this green when it comes to encoding videos and such. But this is something I've always been wondering about ever since I first started trying to make webms years ago for the first time. But I feel like it's good to know whether or not the source video file has any actual impact on the output file when encoding or if it can be ignored completely?
>>6269721Generally, don't worry about the source. There are minor considerations if the fps is not an even multiple, like going down to 24fps from 30, but that's getting into the weeds. There are also minor complications if the source is HDR, you can render an HDR webm but there are compatibility issues that cause the colors to look washed out for many people viewing and you may have to color map it to a standard palette.The optimal thing to do is use the original and best quality source available. Making intermediate clips can result in generational loss (the copy-of-a-copy problem).
>>6269721I can understand as I was the same with command lines. Using ffmpeg directly with the command line can get a bit messy and you gotta know a few things, using multiple filters in a filter chain can seem like a mess and you gotta get the order right etc.>What I would like to know is. Does the source video file have any effects on the output file when encoded to a webm or mp4?On the quality, sure. Even if your video is a larger resolution than what your final encode will be, it still helps due to essentially supersampling. You ideally want to start from the highest source quality possible. This is a 4K HDR source that I tonemapped to 203 nits and downscaled to 1920x1080
>>6269685The OP has a script that just iteratively reduces dimensions of a video. Very lightweight.
>>6269968It's an incredibly inefficient caveman approach. You can pre-calculate everything and hit it in one-shot and in higher quality.
>>6269970Tried to get ChatGPT to update the OP script and optimize it for quality more. Used the same Tekken video.
>>6270278https://files.catbox.moe/qhba7s.rarIn case anyone else wants to iterate. ChatGPT is telling me we should use AV1 to really optimize it, but I think 4chan is actually restricting what codecs you can use.
>>6270280Whoa! You don't say!? Maybe next you should ask it if 4k60fps is what you should be doing to anime all day long.
>>6270280IIRC the only allowed formats here are>WebM containing either VP8 or VP9 video and Vorbis or Opus audio>MP4 containing H.264 video with AAC audio