Slightly different "best encode" question

[burcs]

I know this is asked ad nauseum, but all the answers seem to be geared towards "smallest file". I realize starting with a video that's already been compressed with mpeg2 is not ideal, and h.264 is more efficient. I've tried both CRF 18 and 16 and there's no difference in file size. But what if...

1. I don't care about file size
2. I don't care how long it takes

Like is there an "exhaustive 3 pass that will use whatever bitrate it needs for any given frames that will end up virtually identical to the original without sacrificing quality but can still be played on most devices"? By devices I mostly use PC, so that isn't an issue, but I do have Fire TV/sticks, too.

The mpeg2 averages ~5200, one VOB is around ~6400, and I get mpeg2 sucks. But will a 2 pass @ 5200 with h.264 give me a "better" result than the CRF? The CRF averages ~1900. Would a 2 pass @ 2600 give me the same results than 5200? I get h.264 is more "efficient", but I'm wondering if the 2 pass might compress areas the CRF might not, or vice versa.

Anyway, hope this makes sense. I get that everything's a trade-off between speed/quality/file size, so I'm looking at it more as a "thought experiment." I mean I could do a 2 pass @ 8000 if I really wanted to, but realize that's just a complete waste of hard drive space, etc. Thanks.

[Cornucopia]

Everything sucks at too low bitrate, including h264, h265, av1.
(Nearly) everything can look perfect give high enough bitrate, including mpeg2.
Everything encoded lossily will lose quality with a reencode.

You question is still vague - "most devices"? Whose? "virtually identical"? to Whom?

If you have an original, the best quality you can get is from that original.

Scott

[jagabo]

I've tried both CRF 18 and 16 and there's no difference in file size.

That's not possible. You did something wrong.

1. I don't care about file size
2. I don't care how long it takes

Like is there an "exhaustive 3 pass that will use whatever bitrate it needs for any given frames that will end up virtually identical to the original without sacrificing quality but can still be played on most devices"? By devices I mostly use PC, so that isn't an issue, but I do have Fire TV/sticks, too.

CRF 0 is literally lossless. But it will be far larger than your MPEG 2 source and many devices won't play it. With the slow preset at CRF 10 to 12 it's very hard to see any difference between the source and the re-encoded video, even when A/B switching between enlarged still frames. At around CRF 16 to 18 with most material it's hard to see a difference at normal playback speed.

will a 2 pass @ 5200 with h.264 give me a "better" result than the CRF? The CRF averages ~1900.

Yes. How much better depends on the particulars of the source. But if you lower the CRF to a value that gives you 5200 kbps the quality will be essentially the same as the 2 pass encode.

Would a 2 pass @ 2600 give me the same results than 5200?

x264 at 2600 vs. MPEG 2 at 5200? No. But the difference may not be noticeable on casual viewing. It will depend on the source. If you were asking x264 at 2600 vs. x264 at 5200 -- the latter will be better quality. But how much better depends on the particular video.

I'm wondering if the 2 pass might compress areas the CRF might not, or vice versa.

The bitrate distribution will be nearly identical at the same bitrate (and, of course, other settings being the same). In my experience, bitrate peaks are a tiny bit higher with 2-pass, and troughs a tiny bit lower. But it's not noticeable when watching the video.

[burcs]

I've tried both CRF 18 and 16 and there's no difference in file size.

That's not possible. You did something wrong.

That's entirely possible. It's been years since I encoded it the first time. But I'm (mostly) positive I used CRF 18 back then at Medium preset. This time I tried CRF 16 at Slower. But I also noticed in the settings it said it was going to use 6 reference frames using Slower, and Mediainfo only says 5. So I'm sure something went wrong. But they both came out at 1.6 gb.

As these things go, finally got a decent hard drive so figured I'd re-rip some and do a better encode. I'd be fine leaving it as mpeg2 except there's a couple of corrupt frames that nobody else would notice unless I pointed them out, and that won't do. Never noticed them myself the first time, but they are still there. A simple remux is giving me audio sync issues, likely due to the same corrupt frames, but that's a whole 'nother subject and I've already sorted that, so. Thanks again.

[Cornucopia]

All other factors being equal, using crf16 will ALWAYS be larger & better quality than crf18, though you might not notice the difference.

Scott

[jagabo]

I've tried both CRF 18 and 16 and there's no difference in file size.

That's not possible. You did something wrong.

That's entirely possible. It's been years since I encoded it the first time. But I'm (mostly) positive I used CRF 18 back then at Medium preset. This time I tried CRF 16 at Slower. But I also noticed in the settings it said it was going to use 6 reference frames using Slower, and Mediainfo only says 5. So I'm sure something went wrong. But they both came out at 1.6 gb.

You didn't say you used a different preset and (probably) a different version of x264 (and maybe other differences in processing). If you used the same version and the same settings otherwise, a CRF 16 encoding will always deliver a larger file than CRF 18.

A slower preset will deliver a smaller file. A lower CRF will deliver a larger file. The two may have balanced each other out in this particular case. Though typically, the difference in bitrate between CRF 18 and CRF 16 is far larger than the difference between the slow and medium presets.

[Alwyn]

Leave your video as an MPEG 2. When you come across a device that won't play it, then convert it to MP4.