GH4 The Best Way to Convert 4K to 1080

B

Benjobe

Well-known member
I wanted to know if there was any benefit to using After Effects in 32-bit float mode to transcode the original 4K 100Mbps MP4 files into 1080 220Mbps ProRes422HQ files rather than just shrinking the original files by 50% inside the Premiere Pro timeline. I thought that there may be a benefit when color grading. My results seem to suggest otherwise.


I shot the original 4K 24p 100Mbps clip at sunset with the white balance deliberately set to Tungsten to see how the image graded in post. I used Red Giant's Colorista II in Premiere Pro for the grading. Here are the camera settings:


lens: Tokina 11-16mm with Metabones Speedbooster
f-stop: 2
shutter: 1/25th
ISO: 200
white balance: Tungsten
picture profile: Cine-D -5, -3, -5, -2, +0
curve: +5 Shadows / +0 Highlights

Here's the video comparison:

 
Thank you so much....that one simple test has answered quite a few of my own questions I had regarding the gh4....

1.) How do sky gradients at blue hour look with regards to banding?

Pretty damn good in that footage....I cant really spot any problems there.

2.) How does the Tokina 11-16mm lens if used with the original Metabones speedbooster look on the GH4 - and is there any vignetting?

Again it looks good in that footage - sharp and detailed....I was wondering is that at full wide 11mm focal length? I am also not sure whether to wait for the new GH4 specific version of the MSB that has been rumoured as it is meant to be slightly wider FOV and may let a little more light in....might not matter much after viewing that test.

3.) Does the downscale to 1080P help with grading and is it worth using high bit depth After Effects comps to rescale to 1080p?

Cant really pick any difference myself either. I wonder if you graded the original 4K 4:2:0 footage before scaling if it affects IQ much? Anyway looks like you can push a grade pretty far....who needs RAW right ;-) The only difference for me is I will be using 10bit 4:2:2 Cineform transcoding rather than Pro-res.

4.) Can you use a 1/25th shutter on 4K 25P?

If you can do it with 4K 24P I assume the same for 25P.....great!

5.) How is DR in a real world shooting situation like a sunset landscape scene?

No complaints from me with that image....the only minor (and its very minor) issue is some visible noise in the foreground but overall I am very grateful for that test.

Thanks!!!!!!
 
Did you ever get an answer about the noise in the image? I'm sure some Neat Video would help here but just wondering as to the source.
 
Benjobe said:
I wanted to know if there was any benefit to using After Effects in 32-bit float mode to transcode the original 4K 100Mbps MP4 files into 1080 220Mbps ProRes422HQ files rather than just shrinking the original files by 50% inside the Premiere Pro timeline. I thought that there may be a benefit when color grading. My results seem to suggest otherwise...
Click to expand...

That was exactly what I was pondering about. Thanks a lot for this test!
 
Spectra said:
Again it looks good in that footage - sharp and detailed....I was wondering is that at full wide 11mm focal length? I am also not sure whether to wait for the new GH4 specific version of the MSB that has been rumoured as it is meant to be slightly wider FOV and may let a little more light in....might not matter much after viewing that test.
Click to expand...

Yes, the Tokina was at 11mm with no vignetting. I was disturbed by the noise in the grassy areas as well. I think it may be because of the +5 Shadows curve setting. I need to do more tests when work isn't getting in the way.

I'm glad that the video was so helpful to you.
 
SuperSet said:
Did you ever get an answer about the noise in the image? I'm sure some Neat Video would help here but just wondering as to the source.
Click to expand...

I don't have Neat Video but I do have Red Giant's Denoiser II. I tried applying it to a few clips from the GH4 and it totally freaked out, rendering weird digital garbage frames. I've used it before on GH3 footage and it's fine. I don't know if it can't handle the 4K footage, the 200 Mbps 1080 footage, or maybe I need to upgrade from OSX 10.8.5 to Mavericks.

I tried purchasing Neat Video yesterday and the links were failing. I looks like a good program and affordable. I'll try again today.
 
neat video is great!
Benjobe said:
I don't have Neat Video but I do have Red Giant's Denoiser II. I tried applying it to a few clips from the GH4 and it totally freaked out, rendering weird digital garbage frames. I've used it before on GH3 footage and it's fine. I don't know if it can't handle the 4K footage, the 200 Mbps 1080 footage, or maybe I need to upgrade from OSX 10.8.5 to Mavericks.

I tried purchasing Neat Video yesterday and the links were failing. I looks like a good program and affordable. I'll try again today.
Click to expand...
 
Hi

The input codec shouldn't make any difference on the timeline. This is because all grading/effects etc take place not in the codec itself, but on uncompressed video.

So the original files as MP4 on the timeline are decompressed into uncompressed video frames usually 4.4.4, grading takes place, then they are encoded into your chosen output.

The ProRes files on the timeline are decompresed into uncompressed video frames as well, grading takes place, then they are encoded into your chosen output. End result is identical more or less.

In theory the ProRes is further away from the originals and has more losses introduced as the video content has had an additional decompression/lossy recompression to ProRes before landing on the timeline, but ProRes is visually lossless so it isn't noticable.

What usually hurts is output to more lossy codecs such as H264 4.2.0, then that edited, encoded back to the same, used elsewhere and so on, this is because errors get introduced with colour space conversions and of course it is very lossy.

Regards

Phil
 
Phil_D said:
The input codec shouldn't make any difference on the timeline. This is because all grading/effects etc take place not in the codec itself, but on uncompressed video.
Click to expand...

It sounds to me like you're saying that it's best to edit in the native codec (acquired codec). I understand that to be true for 1080 footage, but I was interested to know if it mattered for 4K footage "shrunk" down to 1080 footage.

I thought that by using After Effects in 32-bit float mode to transcode the original 4K 8-bit 4:2:0 footage into 1080 10-bit 4:2:2 footage that I'd be improving the quality of the footage.

It sounds to me like a good test would be to shoot some footage at 4K 8-bit 4:2:0 and then shoot the same footage at 1080p recorded on an external recorder (like the Ninja Blade) capable of 10-bit 4:2:2. Then "shrink" the 4K footage into 1080 and compare the two different recordings. Has anyone done such a comparison yet?
 
Benjobe said:
It sounds to me like a good test would be to shoot some footage at 4K 8-bit 4:2:0 and then shoot the same footage at 1080p recorded on an external recorder (like the Ninja Blade) capable of 10-bit 4:2:2. Then "shrink" the 4K footage into 1080 and compare the two different recordings. Has anyone done such a comparison yet?
Click to expand...

If no one answers before I do...

GH4 Awaiting my collection on Monday, I have a Ninja2 so will be doing some comparisons of 10bit vs 8bit and the like....
 
You must log in or register to reply here.
Back
Top