h265 vs h264 file size really smaller?

[Bramavic2]

Has anyone compared file sizes of same config video? I've heard h265 should be smaller, but from my test footage they seemed almost the same (4k-30fps). Did anyone have different results?

 

[SkyeHigh]

h.265 can produce considerably smaller files than h.264 of the same clip with a noticeable improvement in detail. However it is content specific and compression effectiveness will depend on the content of the video clip(s). Areas of even colour or texture can be more efficiently compressed.

 

[RogerDH]

H.265 should be able to do 4K 60fps with about the same file sizes as H.264 30fps. Are you sure both are 30fps? What does the EXIF file show for data rates?

 

[Kilrah]

The improvement in efficiency can be either used to make the size smaller, the quality better or a middle ground between both - DJI have decided to use the same size (but thus potentially improve quality).

 

[Bramavic2]

H.265 should be able to do 4K 60fps with about the same file sizes as H.264 30fps. Are you sure both are 30fps? What does the EXIF file show for data rates?

Is Exif data rates somewhat accurate? I was actually trying to do a normal/D-Cinelike comparison test, and my 4k h264 30fps cinelike shows 87mbps on exif, while the same config non-cinelike is showing 92mbps. It doesn't make much sense. In mediainfo both show the same 93mbps.

The same but h265 also shows the same in mediainfo (91mbps in exif).

I used LosslessCut to cut all clips to the exact same length (5s) and the all of them became 56MB files (slightly different exact numbers).

And BTW, the 4k h265 60fps became a 93MB file, with media info showing 153 mbps.

 

[RogerDH]

Is Exif data rates somewhat accurate? I was actually trying to do a normal/D-Cinelike comparison test, and my 4k h264 30fps cinelike shows 87mbps on exif, while the same config non-cinelike is showing 92mbps. It doesn't make much sense. In mediainfo both show the same 93mbps.

The same but h265 also shows the same in mediainfo (91mbps in exif).

I used LosslessCut to cut all clips to the exact same length (5s) and the all of them became 56MB files (slightly different exact numbers).

And BTW, the 4k h265 60fps became a 93MB file, with media info showing 153 mbps.

Sorry, didn't mean EXIF (still not finished first cup of coffee); it's in the MP4 or MOV headers. Anyway, if H.264 and H.265 are encoded with the same bitrates, then as (EDIT) Kilrah said, the H.265 will be higher quality. Does the Mini 3 provide any way to specify video quality?

 

[danwnz]

I would check the ISO between videos as well, higher ISO = more noise = larger files

 

[Kilrah]

higher ISO = more noise = larger files

No, target bitrate is constant.

 

[tuxon64]

I've tried H.264 and H.265 with normal and Cinelike mode and didn't notice any difference in file size. Bitrate also remains about the same at 90-100 mbits. I thought it was suppose to be closer to 150mbits?

 

[Eagle Eye 62]

H. 265 is 10bit VS H. 264 at 8bit. There is more data in an H. 265 file of the same frame rate.

 

[RogerDH]

H. 265 is 10bit VS H. 264 at 8bit. There is more data in an H. 265 file of the same frame rate.

Only D-Cinelike mode is 10-bit, and it can be encoded in either H.264 or H.265.

 

[Eagle Eye 62]

While it is true about D-cinelike, you are wrong about H. 264. It is an 8bit codec and does not support 10bit.

 

[RogerDH]

While it is true about D-cinelike, you are wrong about H. 264. It is an 8bit codec and does not support 10bit.

The H.264 codec certainly supports 10-bit video. Are you saying the Mini 3 doesn't?

 

[Eagle Eye 62]

What I am saying is if you want 10bit video on the Mini3 you need to use H.265 as the H.264 codec will not support it.

 

[RogerDH]

What I am saying is if you want 10bit video on the Mini3 you need to use H.265 as the H.264 codec will not support it.

So, you're saying DJI's (or really, Ambarella's) implementation doesn't support the H.264 "High 422" profile? Perhaps you're right, but I'd be surprised, and I can't seem to find any reference to confirm or deny that.