Same crf18 for x264 and x265, means same quality?

[tijgert]

I'm trying to get a real good file size reduction with the absolute best quality possible, using either x264 or x265 and I don't care how long it takes.

So my question is, all other settings being optimal, does the crf value practically guarantee the same image quality for both x264 and x265? (very slow encode).

Someone got a 28GB movie down to 7.4GB on x264 and 6.2GB on x265, both using crf18, but I can't compare the quality myself.
It seems to me that at crf18 both images should be the best possible for that setting, but would it be near identical?

The reason I ask is ecause I also read that x265 mainly shines in low bitrates, but when there's loads of bits to distribute like with crf18 that x264 still has the superior image quality.

Would crf18 x265 make sense over crf18 x264? Or even crf16.

[microchip8]

The crf values of x264 and x265 do not map 1:1. A crf of 18 in x264 would roughly correspond to a crf of 20 or 21 in x265. Keep in mind that x265 has many more tools to compress the image boosting its quality and lowering file sizes

[RanmaCanada]

They don't map 1:1 and it also depends on the content you are encoding. If it is grainy, x264 is still the champ. Same if it's interlaced. It's cheaper to buy new hard drive space than to take the power and time to convert, in the long run.

[benwaggoner]

They don't map 1:1 and it also depends on the content you are encoding. If it is grainy, x264 is still the champ. Same if it's interlaced. It's cheaper to buy new hard drive space than to take the power and time to convert, in the long run.
Yeah. x265's --tune grain does pretty well in retaining grain, but it increases the bitrate a lot and thus doesn't save much bitrate versus an equivalent x264 encode.

If you don't care about matching the original grain pattern exactly, some --nr-inter can help in reducing bitrate and false-motion artifacts in grain while still retaining grain texture. It applies an adaptive deadzone like process only on predicted blocks, so the noise in references I-frames or intra-blocks generally gets retained, just without moving as much.