Subme 11

[nibus]

I understand that subme 10 gives about a 1-2% improvement over subme 9. So what kind of gains does the new subme 11 give? Or is it considered a placebo option at this point?

Or maybe it was implemented so we can say it "goes to 11"...

[Target Practice]

the git says a reason to add was:

"Necessary for a future trellis mode decision/motion estimation patch."

it is used only with the placebo preset, so that answers that, no?

[Dark Shikari]

Probably 1-2% at most...

[STaRGaZeR]

I wonder why do you guys call options and a preset placebo, when they do improve compression. Placebo would be x% slower with no compression benefit whatsoever. If there is benefit, then it's not placebo, more like absurdly-slow-and-usually-not-worth-it-over-faster-presets or something.

[LoRd_MuldeR]

Well, if a setting will significantly slow down the encoding process while the improvement is in a range where the difference will hardly be visible (if at all!), then I think "Placebo" is the correct term :)

[STaRGaZeR]

The improvement was there the last time I tested it, and nothing has changed apparently. Again, if there is improvement then it's not placebo. For example I do 1, maybe 2 encodes over a year, and I don't care at all if it takes more time. However, I do care if the resulting encode is 100~200MB smaller, when producing huge files ;)

[LoRd_MuldeR]

Again, if there is improvement then it's not placebo

If there is an improvement that (under normal circumstances) would justify the extra encoding time, I agree.

But if there only is a minimal improvement (if at all), which can be measured but is far too small to be visible, then it is a placebo.

(i.e. it makes no sense using, unless you don't care about encoding time anyway)

However, I do care if the resulting encode is 100~200MB smaller, when producing huge files

I would assume that if you are using 2-Pass mode and decrease (or increase) the target size (target average bitrate) by 1 or 2 percent with identical settings, you won't be able to see the difference...

[nevcairiel]

If you dont care how long it takes, just encode with the placebo preset and be done with it. :)

[STaRGaZeR]

If there is an improvement that (under normal circumstances) would justify the extra encoding time, I agree.

Normal circumstances? What's that? The circumstances you encode in? :p

But if there only is a minimal improvement (if at all), which can be measured but is far too small to be visible, then it is a placebo.

(i.e. it makes no sense using, unless you don't care about encoding time anyway)

If there is any improvement, then it's not placebo. If it makes sense or not, depends on the situation.

I would assume that if you are using 2-Pass mode and decrease (or increase) the target size (target average bitrate) by 1 or 2 percent with identical settings, you won't be able to see the difference...

Nope, crf in the 13-15 range, sometimes lower, blu-ray compatible settings. Depending on the content, length, etc. it varies a lot, but I have seen 500MB saves in my encodes compared to what you would consider non placebo settings, with the same visual quality. Any option that improves compression efficiency is welcome, within acceptable limits of course. 1 month encoding time is not acceptable for me, but double or triple the time of faster settings is.

[Dark Shikari]

"placebo" is a name. It's there to make fun of people who spend way too long encoding their videos. Nothing more.

If you don't like the name, you can change the code in your x264 version to give it a different one. Or you can use the numerical presets.

[STaRGaZeR]

Gotcha.

[burfadel]

Yeah I believe the standard settings are a little conservative, but thats just be :)

[Logan9778]

Placebo is the CORRECT term for Placebo. 4x slower than Very Slow, with results either not in the range of human sight, or clinically insignificant. Adding a few settings, like --bframes 16, --merange 32, --no-fast-pskip, and --aq-mode 2, to the very slow preset, makes it just as good as Placebo, without the 4x loss in speed.

"placebo" is a name. It's there to make fun of people who spend way too long encoding their videos. Nothing more.

Yes, after much painful experimentation, this is exactly what I discovered. Still trying to figure out exactly what benefits --subme 11 might have, if any.