2025 Dolby Vision 4k UHD procedure?

[Z2697]

It is less than 1.5 fps which is an upper guess by me. rect and amp have a much higher penalty than subme 7. And as I said, all looks stunning here. There's more than one way of doing things :)

(high input, MODERATE output, which is also called refinement!)

It's about 12% slower in my testing.
I'm not trying to dismiss your refinement idea but your result looks stunning because it (will) still look stunning without these 3 options.
I'll spend the refinement power on something else. (probably also not very noticable)

[microchip8]

It's about the almost perfect balance between compression/quality and speed. I've refined my settings over the past year that I'm quite satisfied with them for both quality/compression-wise and speed. I can probably lower subme 7 to 5, but won't go any lower or to the default of 2. And yes, I use rect + amp with limit-modes enabled. Without the latter, it's dog slow :(

[GeoffreyA]

What is the consensus on aq-mode for live-action SDR x265? I found that with anime, aq-mode 3 helped with banding in dark areas.

[microchip8]

I swear by AQ 1. At times, I found that AQ above 1 sometimes smears patches of the image, especially in scenes with low(er) energy, like darker ones. Cannot speak for anime as I don't watch/encode it.

[Z2697]

Banding shouldn't be the factor here, I know AQ is created to help with "flat" area, but still...
If you are encoding 10bits (which I'm think everyone should be doing) and still getting obvious banding, that's not a good sign.
But heck, most of the anime stuff "source" you'll find is not delivering 10bits, so some problem may come from the souce.
Anyway, I also prefer AQ1. I felt sad about the fact that "everyone" is developing AQ2 variants. I even wrote a mod that "combined" AQ1 with some other modes. :o

However here's a notice for OP: if you are using hdr10-opt now, you probably wanna give "dark biased AQ" (aq-mode 3) a try, because hdr10-opt is a fixed QP offset based on average luma code value for each block, and it assigns positive offsets (lower quality) to "dark" blocks.
Darker than "0.5 signal level", IIRC, the darker the block, the lower the quality. (0.5 level is 100nits or so... it's not even actually dark.)
Unfortunately, if we know one thing about HDR movies, it's that they actually consist of a lot "dark" blocks.
So an AQ that's biased to drak blocks might be helpful.
Why not just don't use hdr10-opt?
I mean, we still want the negative offset (higher quality) for bright blocks... right?

[microchip8]

There's a scene in Blade Runner 2049 where in the beginning, Sapper goes in his yellow house and then washes his hands. When he turns to officer KD/Joe and looks at him, in the top-most right corner I could clearly see smearing by AQ 2. I did not try AQ 3 or 4 (it's a dark scene so these may help), but when doing the same encode sample with identical settings except this time with AQ 1, the smearing was completely gone. Like, completely! This is just but one example where I've caught AQ 2 to smear small patches of the image. Further, AQ 1 makes noise less pulsating. I wrote that in another thread but the guy said it doesn't preserve noise better than others. I didn't said preserving noise better (it doesn't), I said it makes it less pulsating.

Sometimes the more simple algorithm is the better one :)

[GeoffreyA]

But heck, most of the anime stuff "source" you'll find is not delivering 10bits, so some problem may come from the souce.

Yes, the art of debanding anime is a field unto itself, and most are problematic at the source. Steins;Gate is a lovely example of copious banding.

There's a scene in Blade Runner 2049 where in the beginning, Sapper goes in his yellow house and then washes his hands. When he turns to officer KD/Joe and looks at him, in the top-most right corner I could clearly see smearing by AQ 2. I did not try AQ 3 or 4 (it's a dark scene so these may help), but when doing the same encode sample with identical settings except this time with AQ 1, the smearing was completely gone.

BR2049 having little to no grain, perhaps smearing would show up more easily than with grainier material?

[microchip8]

BR2049 has so-called mosquito noise which you can see in darker scenes.

[GeoffreyA]

I swear by AQ 1. At times, I found that AQ above 1 sometimes smears patches of the image, especially in scenes with low(er) energy, like darker ones. Cannot speak for anime as I don't watch/encode it.

Anyway, I also prefer AQ1

Thanks. Yesterday, I encoded Last Night in Soho using AQ1 and am very pleased with the quality. Though I haven't compared with the other AQ modes.

[VoodooFX]

What is the consensus on aq-mode for live-action SDR x265? I found that with anime, aq-mode 3 helped with banding in dark areas.

AQ 1, as has already been suggested. If you've problems with the dark areas, try increasing the strength to 1.1–1.2. Maybe 1.1 would be a good overall default.