But in terms of x264 - quality is set by the CRF value. CRF-19 will always produce the same quality (roughly speaking). With slower values (the so-called right values) it will produce the same quality as the “fastest” preset but at a (slightly) lower bitrate. You can’t feed settings into x264 that will magically make CRF-19 look like CRF-16.
^ This is absolutely true, You can’t feed settings into x264 that will magically make CRF-19 look like CRF-16. However, you can choose to do a multi-pass VBR encode INSTEAD of using a CRF value. If we choose the exact same average bit-rate as CRF-16, the file size will of course be the same, and the quality (according to the documentation) will probably not be visibly or measurable better. But, if CRF 17 produces a 21 GB file and CRF 16 produces a 25 GB file, then a 2 Pass encode could be targeted to produce a 23 GB file which would still fit on a single layer disc, but would be slightly higher quality than CRF 17 - call it CRF 16.5. Also, there are many other x264 options that can be adjusted from the defaults which may improve picture quality at the same file size. If anyone knows what they might be, we’re open to suggestions. Jan, I’m testing your suggestion out right now.
I will test as many options as I can (and more than one encoder) on a short section of the film, including what it might look like if it was a full BD-50 (not that we are likely to go that route - unless the 25 GB vs 50 GB comparison is like night and day), and post the results somewhere.
Yes you’re absolutely right, you can use 2-pass to target BD-R size if desired. You can also “hack” the rate by doing two tests on what you think is an “average” section of the movie and extrapolating the data to work out the CRF that will produce the required bitrate/file-size.
That’s fine to do that, all I’m hoping for is that for version 1.5 we get a .mkv with an unrestricted filesize using CRF-16, in addition to the BD25 version. 😃 They can both exist!
The settings used on the current 1080p essentially neuter the release. Even scene groups are required to use better quality presets these days. The motion estimation settings alone cause a huge amount of grain shift from frame to frame. There is so much more to x264 encoding than CRF values and bit rate. I already contacted TM-1 with baseline settings that would dramatically improve their video quality. I just hope they use them.
I wouldn’t say it “neuters” the release. You are right about the per-frame grain using their settings. And that’s not as important with a HDTV source that’s mostly devoid of gain because the quality difference between key frames and others is less obvious, which probably isn’t what you meant with “scene releases” though. I agree with you that you do want each individual frame to look more consistent in quality with its neighbours than the v1.0 release.