SVT-AV1 0.5 Released As Intel's Speedy AV1 Video Encoder

[Spooktra]

"Objective quality benchmarks are algorithms that compare the compressed video with the source and render a value that predicts how the compressed file would fare in subjective tests."

What open source tools would be recommended to perform an objective quality benchmark?

FFMPEG includes tools to measure both SSIM and PSNR; SSIM varies between -1 (completely different) to 1 (exactly the same); SSIM values greater than .99 will usually correlate with pristine quality, .98 will correlate with acceptable quality and .97 will correlate with poor quality; with PSNR it is measured in dB and a measurement of 50 dB or better is considered mastering quality.

If you were to compare an encoded file to the source and calculate both a PSNR of 50 dB or greater AND an SSIM value of .99 or greater, then chances are you would have a hard time telling the encoded version from the source, but note to accomplish such a thing requires very high bit rate even with the best encoders.

[edwaleni]

"Objective quality benchmarks are algorithms that compare the compressed video with the source and render a value that predicts how the compressed file would fare in subjective tests."

What open source tools would be recommended to perform an objective quality benchmark?

[andreano]

three dimensions:

output size × output quality (many automated metrics or perceptual) × encoding speed

To ease visualization you usually equalize one dimension, like trying to configure all encoders to produce same size, or same quality according to some automated metric, and then plot the other two dimensions.

This is not a focus of Phoronix.

Well said, but I know a very practical methodology that might be within Michael's reach (or someone could help add to PTS if he's busy):

For each encoder and bitrate, find the slowest speed setting that is faster than real-time. Now we are down to 2 dimensions; bitrate and quality, which can be measured and plotted.

[wswartzendruber]

If Intel keeps this up, they're going to be a software company.

[jntesteves]

Video encoder quality is hard to assess because there are three dimensions:

output size X output quality (many automated metrics or perceptual) X encoding speed

To ease visualization you usually equalize one dimension, like trying to configure all encoders to produce same size, or same quality according to some automated metric, and then plot the other two dimensions.

This is not a focus of Phoronix.

Edit: note that if one of these three dimensions is missing, a test is meaningless. It doesn't measure anything of practical use.

[Spooktra]

From a quality point of view, SVT-AV1's output is comparable to that of x264 (source: https://forum.doom9.org/showthread.p...39#post1871939 ), and I doubt that it is as fast as the latter...

I don't know what's sadder, the fact that you didn't bother to read what you linked to, the fact that what you did read you didn't understand or the fact that someone actually "liked" your post.

For those that don't know, here is the command line used for x264 in that test:

x264 --preset veryslow --tune ssim --crf 16 -o test.x264.crf16.264 orig.i420.y4m

For the uninitiated, the "very slow" preset is considered the "mastering" quality preset and crf 15 is generally considered visually lossless to the source; for this test they used "very slow" and crf 16 which means it would be nearly impossible for the average person to tell the difference between the source and the encoded version.

Further, the "very slow" preset, as the name implies is indeed very slow, encoding times are glacial.

If SVT-AV1 matched x264+veryslow+crf 16 with -q 20 -enc-mode 3 (as the command line says) then it doesn't get much better than that because with settings as aggressive as x264's SVT-AV1 would smoke it.

I would say that Intel's SVT family of encoders are the future, the only limitation I can see, if you want to call it that, is the high ram requirements, using ffmpeg patched to support SVT-HEVC, 4k encoding requires nearly 9gb of ram, on my R5 1600 with 8GB ram I can't do it, but on my 4790 based Xeon with 16 GB of ram it encodes just fine.

Here's some good info:

https://01.org/sites/default/files/d...sual_cloud.pdf

[HadrienG]

From a quality point of view, SVT-AV1's output is comparable to that of x264 (source: https://forum.doom9.org/showthread.p...39#post1871939 ), and I doubt that it is as fast as the latter...

[Go_Vulkan]

Is it true that this encoder produces poor quality results?

An important point, which has not been covered by Phoronix yet. Instead, speed seems to be the only metric.

[quikee]

Is it true that this encoder produces poor quality results?

SVT-AV1 had the speed because it didn't implement all the coding tools AV1 has to offer and therefor the quality wasn't nearly as good as aomenc. With new versions they are adding new coding tools so the quality will increase, but probably speed will suffer.

[sophisticles]

Is it true that this encoder produces poor quality results?

No. I have tested the SVT-HEVC encoder here, with samples:

https://forum.videohelp.com/threads/...C-encoder-test