XAVC-S vs XAVC-HS - What situations is one better?

Now I had assumed:

HS = H265 = Smaller file sizes / better image quality, due to the more efficient compression, but takes more computer power to edit the files because H265 is more intensive.

S = H264 = Slightly larger file sizes / less retained details (marginal probably), but a bit easier to edit.

S-I = Huge file size, great details, easy on the PC.

I only have a V60 sd card atm so ignore S-I as an option. No problem there.

Because my PC is overkill personified, I figured HS would be a good choice. Smaller files for the sd card, and can just decode or convert them on the pc later.

Did some tests though... and the file sizes end up being the same in most cases, and sometimes the S files even end up smaller (which confused the heck out of me).

So now I'm wondering if it's a detail thing, such as 100mbps on HS gives the same quality retention as 200mbps on S, or some such thing (which is perfectly possible). However it's hard for me to do a decent test case for that right now.

So I figured I'd try asking here, as these comparisons have been around for at least a year or two due to the A7iv and A7Siii releases, so hopefully there's a comparison or something out there (I've tried googling and youtubing but found nothing, I suspect I'm searching for the wrong keywords though).

Appreciate any pointers!

Nirurin said:

Now I had assumed:

HS = H265 = Smaller file sizes / better image quality, due to the more efficient compression, but takes more computer power to edit the files because H265 is more intensive.

S = H264 = Slightly larger file sizes / less retained details (marginal probably), but a bit easier to edit.

S-I = Huge file size, great details, easy on the PC.

I only have a V60 sd card atm so ignore S-I as an option. No problem there.

Because my PC is overkill personified, I figured HS would be a good choice. Smaller files for the sd card, and can just decode or convert them on the pc later.

Did some tests though... and the file sizes end up being the same in most cases, and sometimes the S files even end up smaller (which confused the heck out of me).

So now I'm wondering if it's a detail thing, such as 100mbps on HS gives the same quality retention as 200mbps on S, or some such thing (which is perfectly possible). However it's hard for me to do a decent test case for that right now.

So I figured I'd try asking here, as these comparisons have been around for at least a year or two due to the A7iv and A7Siii releases, so hopefully there's a comparison or something out there (I've tried googling and youtubing but found nothing, I suspect I'm searching for the wrong keywords though).

Appreciate any pointers!

Click to expand...

The answer is it depends

Interceptor121 said:

Nirurin said:

Now I had assumed:

HS = H265 = Smaller file sizes / better image quality, due to the more efficient compression, but takes more computer power to edit the files because H265 is more intensive.

S = H264 = Slightly larger file sizes / less retained details (marginal probably), but a bit easier to edit.

S-I = Huge file size, great details, easy on the PC.

I only have a V60 sd card atm so ignore S-I as an option. No problem there.

Because my PC is overkill personified, I figured HS would be a good choice. Smaller files for the sd card, and can just decode or convert them on the pc later.

Did some tests though... and the file sizes end up being the same in most cases, and sometimes the S files even end up smaller (which confused the heck out of me).

So now I'm wondering if it's a detail thing, such as 100mbps on HS gives the same quality retention as 200mbps on S, or some such thing (which is perfectly possible). However it's hard for me to do a decent test case for that right now.

So I figured I'd try asking here, as these comparisons have been around for at least a year or two due to the A7iv and A7Siii releases, so hopefully there's a comparison or something out there (I've tried googling and youtubing but found nothing, I suspect I'm searching for the wrong keywords though).

Appreciate any pointers!

Click to expand...

https://interceptor121.com/2022/12/05/sony-a1-a7s3-video-codecs/

The answer is it depends

Click to expand...

Haha, the answer I was worried it would be! But thanks for the link I'll have a read through

Interceptor121 said:

https://interceptor121.com/2022/12/05/sony-a1-a7s3-video-codecs/

The answer is it depends

Click to expand...

Weirdly this seems to say that H265 is easier to edit than H264... which I'm pretty sure is wrong? So now I'm not sure I can trust any of the other conclusions they draw, other than the noise being the same / similar enough (which is good).

Nirurin said:

Interceptor121 said:

https://interceptor121.com/2022/12/05/sony-a1-a7s3-video-codecs/

The answer is it depends

Click to expand...

Weirdly this seems to say that H265 is easier to edit than H264... which I'm pretty sure is wrong? So now I'm not sure I can trust any of the other conclusions they draw, other than the noise being the same / similar enough (which is good).

Click to expand...

Nope it depends on the subsampling

if you choose 420 hevc is accelerated

if you choose 8 bits h264 is accelerated

however nobody would choose either of the above and hevc 422 is slightly faster than h264 as the implementation is very complex

Obviously if you don’t understand any of this it will sound alien to you lol

file size between Hs and S doesn’t change so the more or less space is not a point

Try it yourself record the same scene from the camera with the two codecs 10 bits 422 where they are both not hardware accelerated and try a playback. You will see that H264 will stutter and the other one after buffering will play. The GOP structure of the H264 files is complex and it has CABAC the HEVC in comparision are easier to digest. In addition on a 4K file you can't tell them apart either on a static scene.

HEVC works better with larger frame size i.e. 8K with 4K depending on the scene you won't necessarily see a benefit

Also your assumption on level of detail being related to the bitrate are largely incorrect

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/
If you want to get in touch don't send me a PM rather contact me directly at my website/social media

Nirurin said:

Interceptor121 said:

https://interceptor121.com/2022/12/05/sony-a1-a7s3-video-codecs/

The answer is it depends

Click to expand...

Weirdly this seems to say that H265 is easier to edit than H264... which I'm pretty sure is wrong?

Click to expand...

it is wrong, and if you want actual facts you should ignore the constant link spamming that he posts to his own website, for instance see this post in a thread that he started: Re: Quad Bayer Mobile Sensor Technology in Full Frame Camera?: Photographic Science and Technology Forum: Digital Photography Review (dpreview.com)

Nirurin said:

So now I'm wondering if it's a detail thing, such as 100mbps on HS gives the same quality retention as 200mbps on S, or some such thing (which is perfectly possible).

Click to expand...

that is generally correct, you'll have to use more bandwidth on h.264 to get similar p.q. to h.265.

you can get h.265 hardware decoding acceleration with late-model intel cpu's, and the m1 mac, it doesn't necessarily have to be harder.

MILC man said:

Nirurin said:

Interceptor121 said:

https://interceptor121.com/2022/12/05/sony-a1-a7s3-video-codecs/

The answer is it depends

Click to expand...

Weirdly this seems to say that H265 is easier to edit than H264... which I'm pretty sure is wrong?

Click to expand...

it is wrong, and if you want actual facts you should ignore the constant link spamming that he posts to his own website, for instance see this post in a thread that he started: Re: Quad Bayer Mobile Sensor Technology in Full Frame Camera?: Photographic Science and Technology Forum: Digital Photography Review (dpreview.com)

Click to expand...

It has not yet been demonstrated who is right or wrong actually on that specific point despite the long essays of the nah sayers

MILC man said:

Nirurin said:

So now I'm wondering if it's a detail thing, such as 100mbps on HS gives the same quality retention as 200mbps on S, or some such thing (which is perfectly possible).

Click to expand...

that is generally correct, you'll have to use more bandwidth on h.264 to get similar p.q. to h.265.

Click to expand...

Not necessarily in this specific case the HEVC implementation is very light and not particularly efficient. The I frames are marginally bigger and very few majority are bidirectional predictions from a rare reference frame. H264 has 2 reference frames and CABAC the implementation is extremely tuned and visually it is almost impossible to take them apart. If you had tried you would know

MILC man said:

you can get h.265 hardware decoding acceleration with late-model intel cpu's, and the m1 mac, it doesn't necessarily have to be harder.

Click to expand...

Acceleration is limited to HEVC 420 subsampling because this is the standard for internet playback. H264 acceleration was for 8 bits as that is an older codec.

The camera makes files that go all the way to 422 which is not accelerated in any scenario those are the one that make the computer choke

The way the H264 is implemented in the Sony cameras is more taxing than HEVC which is a very basic implementation with a very long gop and not very compressed at all considering what HEVC can do. The only reason why HEVC is introduced is because H264 is unsuitable to 8K. For encoding even 4K H264 remains the standard at present.

If you know how you can use ffprobe to see what is inside the gop structure instead of putting out wild guess or generic statements

It is apparent that here on dpreview the general understanding of video is extremely limited so maybe the op should go elsewhere to get some other perspective

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/
If you want to get in touch don't send me a PM rather contact me directly at my website/social media

Technically speaking, h.265 is supposed to be a lot better than h.264. In most peoples' experiences, h.265 has not lived up to the hype in terms of a format improvement. Many find h.265 more difficult to edit and the image quality improvement is just not that visible.

I do a lot of video and frankly, just stick with 10 bit 4:2:2 h.264 (XAVC-S) and you will make beautiful images. Easy. Done. No need to keep up with the current tech trend and your files will work everywhere.

But please do the comparison for yourself as first hand knowledge is the most important thing. I think you will conclude that there is very little difference between the two formats.

Bassman2003 said:

Technically speaking, h.265 is supposed to be a lot better than h.264. In most peoples' experiences, h.265 has not lived up to the hype in terms of a format improvement. Many find h.265 more difficult to edit and the image quality improvement is just not that visible.

I do a lot of video and frankly, just stick with 10 bit 4:2:2 h.264 (XAVC-S) and you will make beautiful images. Easy. Done. No need to keep up with the current tech trend and your files will work everywhere.

But please do the comparison for yourself as first hand knowledge is the most important thing. I think you will conclude that there is very little difference between the two formats.

Click to expand...

Good advice. You need to think the HEVC implementation that Sony does is very light. There are very few reference frames and many predictions. In essence this is only good for static scenes.

On the decoding though XAVC-HS at 50 fps decodes better than XAVC-S which is instead a full blown high quality implementation

In a way the XAVC-S implementation is so good that the basic XAVC-HS codec does not really offer much benefit for majority of users.

My Mac Mini with Vega64 gpu can't edit real time H264 50/60 fps files from the camera while it manages H265 just fine because they are not that fine tuned

And XAVC-HS does not have a 30 fps mode which pretty much kills it for online use while XAVC-S covers the entire range of frame rates

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/
If you want to get in touch don't send me a PM rather contact me directly at my website/social media

This is why it is good to do research and know the implementation of new tech. Otherwise you might purchase more expensive SD cards and maybe computer upgrades etc... thinking newer must be better. The 10 bit 4:2:2 is more important than the compression standard since h.264 is very mature and not restrictive for most uses. From this perspective, the V60 cards cover most everything and are a great value.

Bassman2003 said:

This is why it is good to do research and know the implementation of new tech. Otherwise you might purchase more expensive SD cards and maybe computer upgrades etc... thinking newer must be better. The 10 bit 4:2:2 is more important than the compression standard since h.264 is very mature and not restrictive for most uses. From this perspective, the V60 cards cover most everything and are a great value.

Click to expand...

It might be better to Shoot in XAVC HS instead of XAVC S-I on the A7IV : SonyAlpha (reddit.com)

This person seems to have done some comparisons and the HS codec shows some improvements in fine details (with regard to the compression artifacts) over S-I. They don't compare to S, but they're both H264 so I'd assume they would be similar.... ?

It's hard for me to test, and tbh it does seem like there's very little difference in either way.

Long GOP compression gets a bad rap but it is actually very efficient where intra-frame codecs are not very efficient. Meaning it takes a lot more bitrate on an intra-frame (all-I) codec to equal a smaller bit rate of a long GOP recording. In short, the 240Mbps might be a bit under for its implementation. Those clips were heavily graded s-LOG footage, so who knows how they were shot etc... All I ever shoot is long GOP. It is time tested and works great. No need to keep searching

Bassman2003 said:

Long GOP compression gets a bad rap but it is actually very efficient where intra-frame codecs are not very efficient. Meaning it takes a lot more bitrate on an intra-frame (all-I) codec to equal a smaller bit rate of a long GOP recording. In short, the 240Mbps might be a bit under for its implementation. Those clips were heavily graded s-LOG footage, so who knows how they were shot etc... All I ever shoot is long GOP. It is time tested and works great. No need to keep searching

Click to expand...

I currently only have (and have no plans on changing) V60 memory cards which I brought over from my old camera, so intra was never really an option, I just found it an interesting point of reference.

The main issue for me is figuring out the better option between H265 and H264.

All things being equal, for the same 200mbps, the H265 video should be higher quality. In fact I think the H265 100mbps should be roughly equal to the H264 200mbps. But I don't know if there's other factors in play with how the codecs have been implemented by sony. And noone seems to have tested it for some reason.

Nirurin said:

Bassman2003 said:

This is why it is good to do research and know the implementation of new tech. Otherwise you might purchase more expensive SD cards and maybe computer upgrades etc... thinking newer must be better. The 10 bit 4:2:2 is more important than the compression standard since h.264 is very mature and not restrictive for most uses. From this perspective, the V60 cards cover most everything and are a great value.

Click to expand...

It might be better to Shoot in XAVC HS instead of XAVC S-I on the A7IV : SonyAlpha (reddit.com)

Click to expand...

a test of a tree that could move with the wind is a pretty bad way to make an assessment

Nirurin said:

This person seems to have done some comparisons and the HS codec shows some improvements in fine details (with regard to the compression artifacts) over S-I. They don't compare to S, but they're both H264 so I'd assume they would be similar.... ?

Click to expand...

No long gop on static scenes has around 15% more data in the I frame

Nirurin said:

It's hard for me to test, and tbh it does seem like there's very little difference in either way.

Click to expand...

why don’t you do your own tests it takes little to see if you can tell or not the differences

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/
If you want to get in touch don't send me a PM rather contact me directly at my website/social media

Interceptor121 said:

Nirurin said:

Bassman2003 said:

This is why it is good to do research and know the implementation of new tech. Otherwise you might purchase more expensive SD cards and maybe computer upgrades etc... thinking newer must be better. The 10 bit 4:2:2 is more important than the compression standard since h.264 is very mature and not restrictive for most uses. From this perspective, the V60 cards cover most everything and are a great value.

Click to expand...

It might be better to Shoot in XAVC HS instead of XAVC S-I on the A7IV : SonyAlpha (reddit.com)

Click to expand...

a test of a tree that could move with the wind is a pretty bad way to make an assessment

Click to expand...

The difference has nothing to do with the tree movements, but with the blocks of data used to make up the sky/tree boundaries and gradients.

Whether there were other things that might cause those issues, I don't know. Nothing I can think of.

Though it may be that the bitrate for the intra version simply isn't high enough for intra.

Nirurin said:

Interceptor121 said:

Nirurin said:

Bassman2003 said:

This is why it is good to do research and know the implementation of new tech. Otherwise you might purchase more expensive SD cards and maybe computer upgrades etc... thinking newer must be better. The 10 bit 4:2:2 is more important than the compression standard since h.264 is very mature and not restrictive for most uses. From this perspective, the V60 cards cover most everything and are a great value.

Click to expand...

It might be better to Shoot in XAVC HS instead of XAVC S-I on the A7IV : SonyAlpha (reddit.com)

Click to expand...

a test of a tree that could move with the wind is a pretty bad way to make an assessment

Click to expand...

The difference has nothing to do with the tree movements, but with the blocks of data used to make up the sky/tree boundaries and gradients.

Click to expand...

how do you know? Shooting at 1/50 you get motion blur immediately

in addition macro blocking doesn’t show there but in the area of the sky where there are gradients

as I said unreliable testing

Nirurin said:

Whether there were other things that might cause those issues, I don't know. Nothing I can think of.

Though it may be that the bitrate for the intra version simply isn't high enough for intra.

Click to expand...

That’s not true either and the differences are scene dependant

intra codecs are for fast motion generally there is little benefit to use them on general purpose single frame rate shots

what are you planning to shoot and which frame rate? That determines your choice

in addition there is no 30 fps mode for HS and 30/60 is the best choice for online sharing

[FRAME]t message repeated 194 times

media_type=video

stream_index=0

key_frame=1

pkt_pts=26000

pkt_pts_time=1.040000

pkt_dts=26000

pkt_dts_time=1.040000

best_effort_timestamp=26000

best_effort_timestamp_time=1.040000

pkt_duration=1000

pkt_duration_time=0.040000

pkt_pos=18129285

pkt_size=2450369

width=3840

height=2160

pix_fmt=yuv422p10le

sample_aspect_ratio=1:1

pict_type=I

coded_picture_number=24

display_picture_number=0

interlaced_frame=0

top_field_first=0

repeat_pict=0

[/FRAME]

That is the size of the I frame on a static shot nothing moving this is around 2x the size of XAVC-SI

So for static / not moving scenes where there is low risk of motion interpolation errors the bitrate of XAVC-S at 250 Mbps is too low you would need 480 Mbps to match it

However if you look at 60 fps the same packet size goes to c.1400000 while XAVC-SI would stay at 1200000 now with higher frame rate it becomes much more important to look at fluid motion as you don't really have static shots and XAVC-S long gop is very hard to edit

So if your default is to shoot 60 fps then XAVC-SI has a purpose

Ultimately the intra codec offered by Sony are largely insufficient, they probably wanted to stay within the limits of a V90 memory card or using CF Express A would generate too much heat.

Hardware permitting for general purpose XAVC-S is the best option

A final remark if you shoot Slog3 you can get quite a lot of information with XAVC-S shooting 60fps however if you use a gamma encoded picture profile the bitrates are overall not too high. So you really need to shoot Slog3 on the A1/A7S3

--
instagram http://instagram.com/interceptor121
My flickr sets http://www.flickr.com/photos/interceptor121/
Youtube channel http://www.youtube.com/interceptor121
Underwater Photo and Video Blog http://interceptor121.com
Deer Photography workshops https://interceptor121.com/2021/09/26/2021-22-deer-photography-workshops-in-woburn/
If you want to get in touch don't send me a PM rather contact me directly at my website/social media

Nirurin said:

The main issue for me is figuring out the better option between H265 and H264.

All things being equal, for the same 200mbps, the H265 video should be higher quality. In fact I think the H265 100mbps should be roughly equal to the H264 200mbps. But I don't know if there's other factors in play with how the codecs have been implemented by sony. And noone seems to have tested it for some reason.

Click to expand...

This is what we are trying to let you know - 10 bit 4:2:2 h.264 is the best overall option. Should be and "is" are two different things. Both of us responding have tested it for our uses. We are suggesting you test it for your uses. Otherwise all you have is our opinions.

If you do not feel you can tell the difference, (that's fine) then learn to tell the difference or let it go and accept the advice.

Bassman2003 said:

Nirurin said:

The main issue for me is figuring out the better option between H265 and H264.

All things being equal, for the same 200mbps, the H265 video should be higher quality. In fact I think the H265 100mbps should be roughly equal to the H264 200mbps. But I don't know if there's other factors in play with how the codecs have been implemented by sony. And noone seems to have tested it for some reason.

Click to expand...

This is what we are trying to let you know - 10 bit 4:2:2 h.264 is the best overall option. Should be and "is" are two different things. Both of us responding have tested it for our uses. We are suggesting you test it for your uses. Otherwise all you have is our opinions.

If you do not feel you can tell the difference, (that's fine) then learn to tell the difference or let it go and accept the advice.

Click to expand...

Lol that's right to be frank XAVC-S is the only option that offer all frame rates

I have yet to find a way to change frame rate using a button instead of navigating menus but ultimately if I am shooting a 24/25/30 fps project I use XAVC-S if I have a 60fps project all along I use XAVC-S

HS mode is for 8K I don't bother using it as it does not have neither a 25 nor a 30 fps so that is basically useless overall. In 8K instead it does which is totally crazy

I suggest to test different codecs with LUTs, color space transformation etc.

I noticed that Sony A74 slog3 422 10bit XAVC-S had some weird artifacts when I added a LUT and did some pretty standard exposure/contrast adjustments. Since then I have been recording all-intra to V90 SD cards.