Shooting flat with a DSLR... good in theory, but not so much practicality?

[HansSteinert]

I love color grading. It's probably my favorite part of post production. So naturally, I loved shooting flat. I used the Philip Bloom settings, and then resorted to cinestyle as soon as that came out.

Last week, I had to shoot a promo event, and the necessary turn-around time was very quick. So, I decided to just shoot with the "standard" profile so I didn't have to grade in post.

The image turned out incredibly clean and sharp compared to my past shoots. I'm assuming this is because all the "coloring" happens within the camera BEFORE the camera does the h264 compression. Which made me wonder...

Perhaps shooting flat with the DSLR codec does more harm than good due to the codec and bitrate? Shooting flat and then grading this flat h264 encoded video file seems to really exaggerate artifacting and noise. Perhaps if we want the best images out of these DSLR cameras, we should try to get as close to our finished look as possible within the camera profile settings?

What do you guys think? Have you guys had similar findings? I shoot with a 60D, mind you.

[HansSteinert]

Here's a frame from the weekend using the standard profile:

http://img689.imageshack.us/img689/6...dnograding.png

[c2thew]

cinestyle is reserved for high end video editing. Although i haven't used cinestyle too much, i've noticed it just adds one more layer of work that can be corrected using the import settings, but if you are just taking walk around shots and want to share them asap, then shooting the standard profile would be best.

[benesotor]

Well for cinema especially, sharpness can spoil the look. Generally over-sharp looking images are common in TV or video, but people after a filmic look will gravitate to the flat-contrast image from cinestyle.

But a major advantage of the cinestyle is that unlike the standard picture settings, the input/output luminance doesn't have a sharp roll-off at the lows/highs, so there's plenty more detail to work with in the shadows and highlights, again more filmic. You can always resharpen if you prefer the more punchy look, but it's always better to start with more information and crush it in post... because you can't do it the other way round!

[HansSteinert]

Originally Posted by c2thew

cinestyle is reserved for high end video editing. Although i haven't used cinestyle too much, i've noticed it just adds one more layer of work that can be corrected using the import settings, but if you are just taking walk around shots and want to share them asap, then shooting the standard profile would be best.

I know, but my argument is that even if you have the time and the experience to color it, the picture quality won't be as good as if you got the same look straight out of the camera. This is because the format can't handle heavy color manipulation. When you do this, it exaggerates noise and artifacting.

[benesotor]

Originally Posted by HansSteinert

I know, but my argument is that even if you have the time and the experience to color it, the picture quality won't be as good as if you got the same look straight out of the camera. This is because the format can't handle heavy color manipulation. When you do this, it exaggerates noise and artifacting.

Maybe you need to look again at the way you colour-correct. Log formats are universally used by the industry, and even from my quick tests I've found I can get nicer results from the cinestyle codec.

[HansSteinert]

Originally Posted by benesotor

Maybe you need to look again at the way you colour-correct. Log formats are universally used by the industry, and even from my quick tests I've found I can get nicer results from the cinestyle codec.

Log isn't the problem, the problem is the h264 compression in 4 2 0 color space with 40mbps bitrate. It creates slight artifacting, and when you do heavy grading over that, the artifacting gets very exaggerated. Professionals grade files that are shot flat in 4 4 4 or 4 2 2 color space, with lossless compression.

It's not my grading that is poor, the look of it is great. It's the artifacting that is evident on the image after grading.

These DSLR cameras work like this:

Captured image -> camera profile -> h264 compression

right now, we're doing

flat image -> h264 compression

and then trying to grade that flat h264 compressed image back to something that's pleasing. h264 is a decent codec, but it was designed as a delivery codec, not as a production codec. The codec gets rid of tons of information in the image to keep the file size small in such a way that our eyes hardly notice. However, the computer will notice this lack of information when trying to manipulate the video's colors. The result is that when you try to grade the h264 file, the image begins to fall apart in artifacting. That's why my proposal is to get as close to your finished look within the camera profile before the h264 compression as possible, because that will get you the best image.

[benesotor]

Originally Posted by HansSteinert

Log isn't the problem, the problem is the h264 compression in 4 2 0 color space with 40mbps bitrate. It creates slight artifacting, and when you do heavy grading over that, the artifacting gets very exaggerated. Professionals grade files that are shot flat in 4 4 4 or 4 2 2 color space, with lossless compression.

It's not my grading that is poor, the look of it is great. It's the artifacting that is evident on the image after grading.

These DSLR cameras work like this:

Captured image -> camera profile -> h264 compression

right now, we're doing

flat image -> h264 compression

and then trying to grade that flat h264 compressed image back to something that's pleasing. h264 is a decent codec, but it was designed as a delivery codec, not as a production codec. The codec gets rid of tons of information in the image to keep the file size small in such a way that our eyes hardly notice. However, the computer will notice this lack of information when trying to manipulate the video's colors. The result is that when you try to grade the h264 file, the image begins to fall apart in artifacting. That's why my proposal is to get as close to your finished look within the camera profile before the h264 compression as possible, because that will get you the best image.

Maybe do what I do... convert to Cineform or another 422 10-bit codec. You get the flat image but behaves much better in colour-correction. I agree editing the h.264 footage as it is isn't a great idea.

[HansSteinert]

Originally Posted by benesotor

Maybe do what I do... convert to Cineform or another 422 10-bit codec. You get the flat image but behaves much better in colour-correction. I agree editing the h.264 footage as it is isn't a great idea.

Converting an h264 file to cineform or any other codec doesn't change much. It's still the same source info: h264. The only difference would be that it's easier for your computer to edit.

[benesotor]

Originally Posted by HansSteinert

Converting an h264 file to cineform or any other codec doesn't change much. It's still the same source info: h264. The only difference would be that it's easier for your computer to edit.

Well it does colour-correct better, that was one of the main insights gained from doing the Zacuto DSLR tests.. that results are noticeably improved when you CC with cineform or prores.
By shooting in pre-processed 8-bit h.264 you've got no hope in re-gaining any detail.

[HansSteinert]

Originally Posted by benesotor

Well it does colour-correct better, that was one of the main insights gained from doing the Zacuto DSLR tests.. that results are noticeably improved when you CC with cineform or prores.
By shooting in pre-processed 8-bit h.264 you've got no hope in re-gaining any detail.

The only reason it color corrects "better" is because the project file that you're grading in defaults to 16 or 32 bit I believe, since that's the format cineform is. You can achieve the same thing by manually going into your project with the h264 files and changing it from 8 to 16 or 32 (which I definitely recommend).

When you take an h264 and convert to a lossless production codec, you aren't gaining any information or detail in the video file. It's just easier on your computer to edit because h264 needs to be decoded, whereas cineform is optimized for editing speed.

It's the same effect if I were to take an mp3 file and convert it to uncompressed wav. There's no gain of information. The source recording is still crappy, and that's all the information you will ever have even if you up-convert unless you artificially interpolate it (which usually has crappy results).

[benesotor]

To be honest, I don't find any of what you talk about in terms of artefacts. There may be slightly more noise in the shadows, but that can be combated by applying a good LUT curve. At least I'll have information there if I need it.
The standard picture style I find just gives over-sharp and over contrasty video, which also exaggerates aliasing which is common especially at 720p.

[HansSteinert]

Originally Posted by benesotor

To be honest, I don't find any of what you talk about in terms of artefacts. There may be slightly more noise in the shadows, but that can be combated by applying a good LUT curve. At least I'll have information there if I need it.
The standard picture style I find just gives over-sharp and over contrasty video, which also exaggerates aliasing which is common especially at 720p.

Yeah, I'm not a huge fan of Standard because of the sharpening... I would reduce that for sure. It's just the general principle of shooting close to how you want your finished look to be rather than shooting flat that I was getting at... at least with DSLR cameras.

[ben_r_]

This sounds a lot like a trying to do RAW style editing on a lossy compressed JPEG image. I was under the impression that what came out of these DSLR cameras was essentially a JPEG movie file, already compressed, post processed a bit and anything but raw video data. So editing that in these manners would be like editing JPEG files instead of RAWs where you are much more limited as to what you can do and still have look good.

[Csae]

If you liked the standard profile enough, you should take a look at the Faithful profile.

I haven't used the standard profile in years. Faithful gives me consistently better results in both stills and short clips.