GH5 "Hidden" H.265 Codec for regular use (normally Anamorphic only) - What do you think?

F

filmguy123

Veteran
"Hidden" H.265 Codec for regular use (normally Anamorphic only) - What do you think?

Saw this video today and thought it was an interesting find/thought:

https://www.youtube.com/watch?v=znLS7h2dSXM

Basically, H.265 is only available shooting open gate (?) in Anamorphic mode. He sets up anamorphic mode, disables some anamorphic features, shoots with normal lenses, and crops/zooms to a standard wide screen aspect ratio in post.

He posits that the H.265 files are much higher quality. And perhaps(?) less image processing applied since it's open gate?

Thoughts?
 
I would like to see a side by side with the normal codec vs the h.265 In my view, codecs are not a weak point of the GH5. So any kind of "hack" that introduces more steps has to be better or why bother?
 
He makes a very interesting presentation and illustrates how you can make two images look alike. The sharpening applied to the H265 was for me a bit excessive, but that's just my opinion. I don't think HEVC is inherently better, in fact I prefer my codec to use less compression not more. The compression I'm referring to is GOP, up to 250 pictures for HEVC which is a lot. The more significant improvement from the 4x3 anamorphic mode is simply having the full sensor 6k image. Once resampled to 4k the differences are between in-camera resampling versus NLE.
 
I'd actually be very interested if there was a way, or if Panasonic would enable, an h.265 codec for slow motion modes.

GH5s with 120-240 in h.265 codec...?? I would think it would look a lot better than the current h.264 which gets stretched way too thin, if the CPU could handle that.
 
Tom Roper said:
He makes a very interesting presentation and illustrates how you can make two images look alike. The sharpening applied to the H265 was for me a bit excessive, but that's just my opinion. I don't think HEVC is inherently better, in fact I prefer my codec to use less compression not more. The compression I'm referring to is GOP, up to 250 pictures for HEVC which is a lot. The more significant improvement from the 4x3 anamorphic mode is simply having the full sensor 6k image. Once resampled to 4k the differences are between in-camera resampling versus NLE.
Click to expand...

Testing HEVC 6K vs. H.264 files on GH5 show me that the first have less sharpening. There is a topic about sharpness processing and we got the same conclusion about 6K mode.
 
I would like to see a comparison of this h.265 profile ( which I think is 200 Mbps ) against the standard 72 Mbps HLG 4K 10-bit 4:2:2 h.265 profile.
 
improser said:
Testing HEVC 6K vs. H.264 files on GH5 show me that the first have less sharpening. There is a topic about sharpness processing and we got the same conclusion about 6K mode.
Click to expand...

Yes, the sharpness I was referring to is the sharpness he was adding back in to equate with the raw still image.
 
I like the 6K option (actually more like 5K) is great, it is my default recording option using V-Log or HLG if I do not use the Atomos.

The only problem is that you absolutely must transcode, but even transcoding to an all-intra of this size will still be tolling on DaVinci Resolve.
 
improser, just to clarify your post, are you saying there is a thread here on DVX discussing 6k mode / open gate (the mode to access this h.265 codec) bypasses internal sharpening?

If so, that be a great way to get a more natural look out of the GH camera... can you shoot this mode to standard codecs or do you have to use h.265?
 
The one down side is I believe you get 4:2:0 color when you shoot this mode. Not really a huge deal as it is at least 10bit.

Makes sense that reducing the amount of processing done in camera and saving it for post could yield potentially better results. The GH5 does not have an OLPF however and it uses its in camera scaling method as a form of OLPF. One could potentially get more moire and aliasing by shooting this way. Scaling in post may not always eliminate it either. Slap a OLPF on the camera and scale in post and I bet the camera will look even closer to the EVA1.

Another workflow I'm interested in is a potential 2.5k 4:4:4 master for VFX work. Even though the h265 format is 4:2:0 once you scale down to 2.5k you end up with something almost identical to 2.5k 4:4:4. The movie Ender's Game was shot on 5k RED cameras and scaled down to 2.5k for all the FX work. Not to bring back lost chroma samples but to reduce the loss of detail from the bayer sensor. The open gate on the GH5 could potentially be used for a very similar workflow where the extra resolution from the 2.5k could be used for tracking, match moving and reframing without a loss of detail. Since the open gate can at least be 10bit to support log and HLG you still have a lot of color information to work with at 2.5k. Yeah a ton of processing but if one is ok with a HD delivery this may be a superior way to go.
 
Very interesting stuff. The GH5 is full of surprises. I'd love to see some comparisons between the two.
 
filmguy123 said:
Thomas, if scaling down to HD / 2K...

Thoughts on 4K 60p 420 8bit VS 1080p 60p 422 10bit?
Click to expand...

Yeah, don’t do it.

8bit is 8bit and the damage is already done for log material. That’s why it is more important that the h265 uses 10bit vs 4:2:2. 4:2:2 isn’t a must but if shooting log or HLG the 10bit is much more important.
 
Scale it down to 4K and you have 4:2:2 again. Not that big of a deal.
The 6K anamorphic mode provides excellent results for greenscreen, since it doesn't do any visible sharpening. It looks really soft compared to 4K footage even, though it has slightly more detail, but matches well with ARRI and RED and Varicam footage.
It struggles with banding and blotchiness when there's lots of noise, since the H265 implementation isn't as mature as the H264 10 Bit codec on this camera.
 
DeeZiD said:
Scale it down to 4K and you have 4:2:2 again. Not that big of a deal.
The 6K anamorphic mode provides excellent results for greenscreen, since it doesn't do any visible sharpening. It looks really soft compared to 4K footage even, though it has slightly more detail, but matches well with ARRI and RED and Varicam footage.
Click to expand...

I'd like to see more on this...

Basically 4k 10-bit 422 footage without internal NR/sharpening... sort of a quasi RAW mode...?
 
Thomas Smet said:
The one down side is I believe you get 4:2:0 color when you shoot this mode. Not really a huge deal as it is at least 10bit.
Click to expand...
Makes sense as 4:2:2 is an asymmetrical format that provides both half-resolution chroma channels on each scanline It was originally used back in the Super-VHS/Betamax era to digitize interlaced video, whose individual fields intrinsically have half as much vertical as horizontal resolution. With interlaced H.264 formats, the macroblocks in each field likewise have halved vertical resolution, and 4:2:2 is a good fit as well. But with non-interlaced video, 4:2:2 should produce asymmetrical chroma aliasing if our eyes had enough chroma acuity to notice.

This asymmetry does not occur with 4:4:4 and 4:2:0 formats, which respectively produce full-resolution and half-resolution chroma channels in both horizontal and vertical dimensions. With RGBG Bayer array sensors, luma is mostly correlated with the G channel, which is given twice the spatial resolution as the R & B chroma channels. Hence, 4:2:0 formats are arguably a more natural fit to Bayer sensors than 4:2:2.
 
Last edited:
Lpowell said:
Makes sense as 4:2:2 is an asymmetrical format that provides both half-resolution chroma channels on each scanline It was originally used back in the Super-VHS/Betamax era to digitize interlaced video, whose individual fields intrinsically have half as much vertical as horizontal resolution. With interlaced H.264 formats, the macroblocks in each field likewise have halved vertical resolution, and 4:2:2 is a good fit as well. But with non-interlaced video, 4:2:2 should produce asymmetrical chroma aliasing if our eyes had enough chroma acuity to notice.

This asymmetry does not occur with 4:4:4 and 4:2:0 formats, which respectively produce full-resolution and half-resolution chroma channels in both horizontal and vertical dimensions. With RGBG Bayer array sensors, luma is mostly correlated with the G channel, which is given twice the spatial resolution as the R & B chroma channels. Hence, 4:2:0 formats are arguably a more natural fit to Bayer sensors than 4:2:2.
Click to expand...
Not sure what all this is supposed to prove, 4:2:2 is better than 4:2:0.
 
You must log in or register to reply here.
Back
Top