Encoding 4k content. What speed preset?

[jlw_4049]

Is it worth it to drop from Medium to Slow? As it takes a 4k movie encode about 24 hours on Medium and about 45 on Slow.

Thanks for the replies!

[Blue_MiSfit]

It's "worth it" to go all the way to placebo if you're running a streaming business where your recurring CDN delivery costs are a huge expense, and being able to use a the slowest presets to reduce bitrate a few percent is worth the one-time compute costs to do this.

For home users, just use the slowest preset you can stand :)

[Boulder]

I'd expect the cost/benefit ratio to be quite low. Medium should already look quite nice, but you could try using slow but set --no-rect to gain some speed with hopefully a negligible quality tradeoff. Slow has some useful additions like a better motion search method and more rate distortion optimization going on.

[jlw_4049]

It's "worth it" to go all the way to placebo if you're running a streaming business where your recurring CDN delivery costs are a huge expense, and being able to use a the slowest presets to reduce bitrate a few percent is worth the one-time compute costs to do this.

For home users, just use the slowest preset you can stand :)

Thanks for the information! I want it to look good, I usually encode at Very Slow on x264. However, 4k/x265 on a 16 threaded processor takes WAY to much time lol.

I'd expect the cost/benefit ratio to be quite low. Medium should already look quite nice, but you could try using slow but set --no-rect to gain some speed with hopefully a negligible quality tradeoff. Slow has some useful additions like a better motion search method and more rate distortion optimization going on.

Thanks for the input :)

[Blue_MiSfit]

Yeah, that's what I generally do with x264 as well.

With x265 I often use 'slow' these days. It's quite good, though if you can stomach 'slower' there's some nice improvements. Honestly x265 is so good it's not really necessary unless you want to shave that last few percent.

[excellentswordfight]

Why do you ask us that? Just look at your encodings and decide if you wanna spend 2x the time for the difference. How can we decide that for you?

I always use at least slow if possible, cause it shows a pretty nice improvements when going from medium and I dont mind the extra time cause its still doable for me. But I dont see much point going much lower then slow, especially with the new preset changes were very slow is now almost as placebo levels of encoding times. I think very slow is 15x times slower then slow now, and you will have to be very bitrate starved to be able to tell the two apart. If you are under crf~20 i wouldnt botter with something like very slow. And if medium isnt looking good enough, well just lower the crf if time prevents you from going slower.

[Nico8583]

So you think medium is a good choice with CRF14~16 for a 4K movie (lower than 18 don't give expected results) ? I'm going to make some tests, and slow / no-sao / no-rect / deblock=-1;-1 seems to be good (I don't know if I must change psyrd). I haven't tried medium, I'm trying slower now just to be sure about difference but perhaps I'll try medium.

[benwaggoner]

As of x265 3.0, slower is the preset that exercises most of HEVC's cool features at a non-punitive encoding time. Below slower, presets can have different kinds of artifacts. Veryslow and Placebo look pretty much like veryslow stylistically, but perhaps just a tiny bit better.

[Nico8583]

I've tried slower and it's very slow for me, perhaps if I change my computer it will be OK but now there is about x3 factor between slow and slower... You let rect enable with slower ?

[chainring]

Your question is somewhat close to one I posed regarding the encoding of 4K sources and whether it's worth the time to encode. Basic answer is, it depends. Is the source material genuine 4K? If yes, then by all means encode at full 4K at whatever preset your heart (eyes, mainly) desire. If the source is not 4K, and is most likely a 2K DCI transfer, then upscaled for the final master, well, I've gone the route of resizing to 1440. 2K is 2048, and somewhere in between is 1440. Crop the bars and you have 2560 by X. A fairly decent compromise. Less time to compress and less space. Pick a higher CRF value and/or preset.

[benwaggoner]

Your question is somewhat close to one I posed regarding the encoding of 4K sources and whether it's worth the time to encode. Basic answer is, it depends. Is the source material genuine 4K? If yes, then by all means encode at full 4K at whatever preset your heart (eyes, mainly) desire. If the source is not 4K, and is most likely a 2K DCI transfer, then upscaled for the final master, well, I've gone the route of resizing to 1440. 2K is 2048, and somewhere in between is 1440. Crop the bars and you have 2560 by X. A fairly decent compromise. Less time to compress and less space. Pick a higher CRF value and/or preset.
Note some older UHD TVs can't play 1440p on-device (e.g. app or from USB) for whatever reason. I don't know that it's any issue with any of the mass market devices, but there were a few 1st gen TVs that couldn't.

Wouldn't matter for HDMI playback.

[hydra3333]

any comment on vs gpu encoding and speeds ?
eg nvidia nvenc with ffmpeg or nvencc with vbr_hq is it "good enough" compared to abovementioned settings ?

[benwaggoner]

any comment on vs gpu encoding and speeds ?
eg nvidia nvenc with ffmpeg or nvencc with vbr_hq is it "good enough" compared to abovementioned settings ?
Bit-for-bit, the GPU/ASIC encoders are going to produce markedly worse encodes than x265 at even medium. If you need real-time performance or can use high bitrates, not problem. But if you are trying to maximize compression efficiency even preset --faster likely is still quite a bit better than the best GPU implementation.