How far can we go with sony H264 compression ? Illustration inside.

B

boyadjy

Active member
Hi everyone,

i want to share my experience with other FS700 users out there regarding post production process so we can share knowledge or tips.

These screen grabs are taken from my last short film. It might give an idea of how far (or how close I would say) we can go with the not so cool sony codec. Below you will find before and after post processing images :

EDIT : flat profiles are ungraded of course.

1_1g.jpg



1_1ug.jpg



1g.jpg



1ug.jpg



3_1g.jpg



3_1ug.jpg



3g.jpg



3ug.jpg


Don't hesitate to participate. I personnaly love this camcorder but I hate this codec. I'm pretty sure i'll buy newly realeased raw modules option to get deeper sampled images to work with.

i'm still happy with the results but it took me some precious time.

Thanks for watching
 
Last edited:
I depends which settings you are using. Regarding that it's only 24Mbps and 4:2:0 8bit the codec is imho outstanding.

Off course you wont get the same headroom for grading as with a native 10bit 422 (or better) recording. In my experience since shooting one year with the camera you have to choose settings which match to your situation. Keep a little space for grading but try to get as close to the look you want. If the first images (the dark ones..) are the ones out of the camera and the flat ones are the ones you wanted: why not shooting with more flat settings instead of putting torture onto yourself and the codec? just my 2 ct
 
Last edited:
Osslund said:
I think H264 is fine when implemented at high bitrate but AVCHD stinks.
Click to expand...

They are the same thing: MPEG-4. H.264 is just a web streaming protocol wrapper so it plays with browsers and software video players over the internet. The only thing that affects what you can do with AVC in post is color sub sampling and bit rate. 4:2:2 color sampling and bitrates over 50mbps, i.e. broadcast grade minimums, give you a lot more data to manipulate. Low bit rate 4:2:0 AVC/H.264 do just fine as final release format codecs, but they suck in post for camera originals if you want room for high quality FX and grading. Even transcoding on import doesn't help much, there is still no data depth in the recorded image to work with.

From my own experiments starting with DNG raw originals, 8-bit encoding is not a significant problem, loss of data from 30:1 or higher compression ratios and reduced chroma sampling is a major problem. By the time you get to 24mbps 4:2:0 AVC/H.264, there is less than 5% of the original image data left to work with compared to an 8-bit 4:4:4 sampled MJPEG original.
 
Last edited:
Clermond said:
I depends which settings you are using. Regarding that it's only 24Mbps and 4:2:0 8bit the codec is imho outstanding.

Off course you wont get the same headroom for grading as with a native 10bit 422 (or better) recording. In my experience since shooting one year with the camera you have to choose settings which match to your situation. Keep a little space for grading but try to get as close to the look you want. If the first images (the dark ones..) are the ones out of the camera and the flat ones are the ones you wanted: why not shooting with more flat settings instead of putting torture onto yourself and the codec? just my 2 ct
Click to expand...

I thought the darker ones were the afters.... because it seems to me the flat ones hold a lot more details.
 
Clermond said:
I depends which settings you are using. Regarding that it's only 24Mbps and 4:2:0 8bit the codec is imho outstanding.

Off course you wont get the same headroom for grading as with a native 10bit 422 (or better) recording. In my experience since shooting one year with the camera you have to choose settings which match to your situation. Keep a little space for grading but try to get as close to the look you want. If the first images (the dark ones..) are the ones out of the camera and the flat ones are the ones you wanted: why not shooting with more flat settings instead of putting torture onto yourself and the codec? just my 2 ct
Click to expand...

Yes of course flat images are ungraded :)
 
TheDingo said:
He's opening up the blacks in post, so the detail hidden in the blacks of the original footage becomes visible after processing.
Click to expand...

Unfortunately, this is just impossible with non raw images. Flat cinegamma profile is fantastic on the FS700. This lets you do whatever you like in post. It's easy to crush blacks with so much detail in dark areas when sooting flat. the opposite way is just impossible when color grading high contrast H264 footage.
I also shoot raw with red cameras and it's a total different world in post production. But I just love this FS700. Image, low light capability, dynamic range, and quick operations above all. Impossible to do with Red one MX.
 
chromeboy007 said:
So the codec is AVCHD?
Click to expand...
Absolutely.
That is why I opened up this thread. To share experiences with this stinky codec. For such a bad codec, results are quite interesting also because of flat profiles (cinegammas are very appealing in this camera)
 
boyadjy said:
Unfortunately, this is just impossible with non raw images.
Click to expand...

So the order of the images you posted above is actually: Post processed image first, followed by Flat original image? ( your description made it sound like it was the other way around )
 
boyadjy said:
Correct. Sorry about that.
Click to expand...

...You're going to hate me, but I think I would prefer to see a less "crushed" blacks look in your finished post processed footage. There lots of nice tone transitions that are completely gone in your post processed footage. ( I would be very curious to see something in-between these two looks, where you've got some of the "punch" from the post processed look without losing all that wonderful shadow gradation )

I know that everyone has their own taste on what they want the finished footage to look like, but I think you may have gone a little too far in crushing the blacks. :-)
 
boyadjy said:
Absolutely.
That is why I opened up this thread. To share experiences with this stinky codec. For such a bad codec, results are quite interesting also because of flat profiles (cinegammas are very appealing in this camera)
Click to expand...

Is it me? I have never thought of AVCHD as "Sony's H264", as other camera companies use AVCHD as their capturing formats as well. Also, I always hate coloring H264 DSLR footage and feel that AVCHD is better to work with.
 
I have to admit I was wondering that you shot these crashed blacks but your description was misleading. ;-)

It's a matter of taste, but as TheDingo wrote above, for my personal taste the shadows are crushed too much - but if you like it it's fine.


I am no expert in picture profiles and I made some comparisions some time ago. I compaired some popular recommended profiles as Noah's settings http://noahyv.com/v2/fs700-picture-profile-guide/ Alister Chapman http://www.xdcam-user.com/forum3/viewtopic.php?f=47&t=1097&p=1647#p1647 , Jan Becker C300 skin tone simulation http://janbeckeronline.com/?p=220 some of AbelCines's settings http://blog.abelcine.com/2012/09/18/sony-fs700-scene-files-from-abelcine/etc.. Noah's highlight proctection is great but as with most of these settings you get quite soon noise in the shadows. It's not a question of Sony's codec, it's the limitation of 4:2:0 in 8Bit. Choose a PP for your needs, dont trust too much that you can fix everything in the post. At least until we have 10bit or RAW
 
Actually, it's web compression because When I do a side by side comparison with my master files, there are much more details in the lowlight range. Guys, everything that is downsampled stinks ! even Jpeg compressed web files, lol

Yesterday, I went to the theatre we'll be showing short film premiere to do technical tests, and it's a completely different feeling on the big screen. Rich details, nice contrast, I can't compare with stills posted on the forum.

Also, I don't know about you guys but shooting film is not like shooting still pictures. I'm a professional photographer as well and post treatment is different because you spend a lot of time looking at a still picture. When you're looking at 24 frames per second images, you have to feel the mood of what the director want to put you in. You never see the pictures like the one I posted in this thread. You get the whole mood and perception, no the pixels. That is why I hate looking at still frame from a film and prefer to see a short extract like few seconds or more if possible :)
 
You must log in or register to reply here.
Back
Top