Vegeta, Youtube will recompress the video anyway, so it makes no sense to use such slow options, you can just use a faster preset coupled with a higher crf
I know, but if I am making an encode, I make it as best as I can afford. TBH, these settings I use them mainly for encoding anime BDs in my server... I put them here as an example and clearly said that.
--bframes 10 forces clip to be encoded with level high 4.0 or 4.1 (depending on resolution) which makes the video potentially unplayable on some devices ( preset veryslow configures bframes to 8 which is more than enough for youtube videos, and you're overriding it to 10)
I think 10 is good and playable by all current devices, even my very old bad laptop. Again, this is something you can modify to suite you... I might choose 8 for faster encodes but there is nothing wrong with 10 and surely I don't go below 8.
setting a preset doesn't mean I want everything from it... it is just a starting point then I modify what I want.
--ref 16 is redundant, because it's set by --preset veryslow. If you use better crf and faster preset (like --preset slow with --ref 5) you'd be fine and get pretty much same quality after youtube recompresses everything . Raising the --ref to 8 is basically the major difference between this profile and --profile "slower".
As I said before, I put the slowest accepted preset then put my settings... If one of them is in the preset, it won't be a problem. for youtube to recompress everything, first you should get a very good quality HD video... then when youtube does its thing on it, it will be better than if you posted a medium quality video in the first place. Ref values helps in that alot.
I can use your suggestion by making it high crf and mid-high ref to make it a very good quality video (IF I CAN AFFORD IT). If not, I stick to ref 8 or something similar.
putting crf 16 is enough for real world videos (as it is perfect for anime). I bet if you choose crf 20 it will also be great but I can afford uploading a big sized video so I go with best settings.
--deblock -1:-1 is potentially a bad choice because it lets x264 slightly blur the image which may be bad if your Youtube videos have lots of screen capture... not to mention that you get the same result just using --tune film ...
this is weird, are you sure? what I know is + deblock blurs the details while - one doesn't. I don't use tunes at all and will never do.
psy-rd 0.75 changed from the default of 1 is again something that makes sense for cinema movies... you're basically telling x264 that it's ok to blur some more and drop grain if it improves compression in frames that are really bitrate starved, which again may not be smart for youtube videos (and pointless because most youtube content is static scenes with Dave talking or showing up something in the case of EEVblog).
I think you are right here, but the value 1 will produce big size video for no gained quality I guess... psych enhancements are not required for youtube videos like you just explained but I can not put it to 0, so yeah, I will make it 1.
-aq-mode 3 is overkill for youtube videos, the default 1 is good enough... youtube videos don't have transitions to dark scenes and the other way around, so that you'd want x264 to take bits from some frames and allocate more to fade outs or darker scenes and so on.. that's what basically aq-mode 2 or 3 does, tweaks the adaptive .
this is your own personal choice to have aq-mode 1, but it is better to have it 2 here. 1 will give you huge file size for nothing gained but 2 will give smaller size with nearly the same quality as there are no motion no grain in it. 1 is not adaptive and only recommended to be used in a high grain video where mode 2 will fail (at least for me).
it is not a crime nor wrong to use it too.
about mode 3, it is the same mode 2 with better handling of dark scenes, so using it all the time instead of 2 is safe (i think) because if it is not a dark scene, it works like 2 or nearly the same. I also said you can replace it by 2 with no problems as I will do.
I didn't check the official build of x264 for a long time, so I will check about this mode 3. last time I checked doom9 about it the posts said that mode 3 and 4 are experimental and won't go in the official build until they aren't.
this is not the only reason I use tmod, it is the mixaq oreaq thing with other patches.
--direct auto is redundant, it's included in all presets higher than medium
xD
For youtube, you can just use --crf as low as you can afford , basically it's up to you and your internet connection, how much you're willing to upload for each minute of encoded video.
for me it is not like that, I do want a perfect result as much as I can... at least for archiving purposes.
BTW, did you try 10-bit videos for youtube? I know it is bad but I want to try it xD. It will be a very long journey until we can use HEVC in youtube, even I still have a bad laptop that can not handle it properly.
and PS. avoid Lagarith, since it's Windows only and relies on floating point implementation which is x86 dependant, some future processors may be different enough that you won't be able to decode your files or your encoded files won't be playable by people using older computers or ARM based systems. MagicYUV (and other lossless codecs like UTVideo or Huffyuv) are much much faster than Lagarith.
good info, thanks!
for me I will use x264 lossless, it is the best for me. I will just install x264 vfw for it with updated tmod version manually put in it.
No, --crf is short for constant rate factor, which is pretty much "constant quality", in a super simplified way it's constant quantizer on steroids... constant quantizer is sort of like treating each frame of the video as a jpeg picture and trying to compress each frame at a fixed quality preset (like jpeg 85%), ignoring how much space that frame will take. Constant Rate Factor also does some psycho-visual stuff, it analyzes each frame and sequences of frames to see where there's darker areas in the frames or stuff that your eyes won't notice easily and moves the saved bits towards the other.. and plays with the underlying qp varying in small steps the quantizer when needed to preserve quality
the term "constant quality" is not achievable through crf only. it is constat rate factor as you pointed out and to get the exact same quality for all frames you have to use "qcomp 1" which is bad. so quality is not gonna be "constant" at all. As you said, the encoder will analyze the video frames and pick the right bits to each frame while not caring about final size.. unlike "bitrate" mode.
So if you call it constant quality, it can be accepted as a simplified term which these guis made it an official name, and people think it is gonna be that simple as their guis tell them.
Actually, the term constant quality applies to using "QP" option rather than "CRF"!
If you choose qp 20 it will always be 20 for all frames but crf 20 is not! it will vary depending on your settings!
_______________
mariush
you must be one of doom9 members like me xD. I just know that.
We had a good discussion, but let's not spam the post with x264 battles xD.