PDA

View Full Version : RAW or ProRes?



Cinemon
07-12-2013, 06:05 AM
hi,
I would like to know what is the best way to get filmic pictures out of BMCC. RAW or ProRes?

I have heard from many that ProRes has a much more cinematic picture than raw, ooc.

To me it's all about the pure image quality. I have enough SSD's and time :D
I have recorded a lot RAW and i get out very gray and lifeless pictures.
Even after adding a lut and try to bring out a radiant image.

What is your opinion?

MWL
07-12-2013, 06:43 AM
You can get cimematic images from either RAW or Prores, RAW is a little more forgiving, but I don't see huge differences in the actual image quality (unless you need to push it around a lot in post).

I think if you could post some DNGs which you consider flat, we could have a go at grading them and see if your perception is changed?

Also just a thought if you are underexposing you might get images which appear flat because of the overall lack of contrast. Feed your camera lots of light.

razz16mm
07-12-2013, 08:24 AM
There is tons more grading room in raw than in Prores. But unless you have been grading raw stills enough to develop the necessary skills to use it effectively the learning curve can be steep. Most of the advantages are in the raw processing side before you transcode to video. You have to spend the time and effort to learn raw processing tools to get the most out of it. So the downside to raw is that there is tons more grading room. Gray washed out images are not inherent to properly graded raw video any more than they are for stills.

http://vimeo.com/58045466

funwithstuff
07-12-2013, 08:37 AM
Both Raw and ProRes can give good results in terms of colour. For Raw you need to know Resolve/AE/Lightroom well; for ProRes you need to know your NLE's tools (or Resolve) especially with regard to contrast. You can push Raw further, but unless you want a really extreme grade or want >2K you'll probably get good results from either. You may wish to use 709 as a starting point rather than Film if you're having trouble with too-flat images.

If you're on FCP X, check out CoreMelt Free for the BMCC DeFlat filter as a good starting point.

http://coremelt.com/free.html

rze
07-12-2013, 08:58 AM
There is tons more grading room in raw than in Prores. But unless you have been grading raw stills enough to develop the necessary skills to use it effectively the learning curve can be steep. Most of the advantages are in the raw processing side before you transcode to video. You have to spend the time and effort to learn raw processing tools to get the most out of it. So the downside to raw is that there is tons more grading room. Gray washed out images are not inherent to properly graded raw video any more than they are for stills.

http://vimeo.com/58045466

Please explain how there is tons more grading room, when both formats have the full luminance range of the sensor. I've tested exposure latitude in the highlights and the darks and saw no real advantage in RAW over prores. As long as you don't clip the sensor you can bring everything back just as easily in prores. RAW is a little safer, but most of that extra security is mental.

If BMC lutted color spaces where not already 90% there for finishing, I would agree with you guys, but BMC rarely needs an intensive grade so the difference is not that large.

Cinemon
07-12-2013, 10:28 AM
Here are some DNG's from me to test with. They look very grey and green tinted...

https://mega.co.nz/#!00ZEgRAS!Hn_7LT9S_wYtrF3eqq12aka0QhwBMQwcFHv93cg-4jE

morgan_moore
07-12-2013, 11:52 AM
I cant see make the download work but the raw processor should have a tint slider to deal with the green and a curve should sort the Grey!

some raw processors dont see where the files are supposed to 'start' which should be in the file meta data

S

Kholi
07-12-2013, 11:56 AM
For anyone having issues with images:

1. Use IR NDs, that's a big problem I'm seeing. Use them at every stop, and unless you're shooting wide open to say F4, use a .3 IR ND as well. Shooting at F11~F16 will likely produce undesired results.

2. Trying to fit 15-16 stops into 13? The camera has a lot of DR, but it isn't THAT much. Unless you REALLY don't want a window or something to clip and you have no choice, let it go, flag, ND, etc.

3. White balance is pretty important, during or after, but white balance.

4. The camera is the camera, it's your job to be cinematic with it.

5. Spend more time in post learning the image, and how to process it. That's as important as learning how to shoot with the camera in this case.

Alán Ortiz
07-12-2013, 12:19 PM
....and if you have some footage that has been properly exposed and are still scratching your head - give Captain Hook's LUT a try. This won't be the answer for every look but its a good place to start if you want some immediate satisfaction. Often times its as simple as apply hook's LUT, finesse the white balance, contrast/levels, and add or subtract saturation to taste - all in one node.

Also - if I had a ton of time, SSD's, hard drive space and processing speed - I would probably shoot everything to raw dng. Not everyone will agree but I percieve a difference in resolution and moire (less in raw dng). Its subtle but its there and always nags me, leading me to believe I should have shot raw in the first place.

razz16mm
07-12-2013, 05:56 PM
Please explain how there is tons more grading room, when both formats have the full luminance range of the sensor. I've tested exposure latitude in the highlights and the darks and saw no real advantage in RAW over prores. As long as you don't clip the sensor you can bring everything back just as easily in prores. RAW is a little safer, but most of that extra security is mental.

If BMC lutted color spaces where not already 90% there for finishing, I would agree with you guys, but BMC rarely needs an intensive grade so the difference is not that large.

Prores is an encoded video format that conforms to display reference standards, REC709 for example. The dynamic range is compressed to meet the defined display gamma of the standard. So within the 0-100 IRE video range as shown on a vectorscope you have a maximum linear DR of 7 stops between black clip and white clip and approximately a 5.5 stop usable range for visible resolved detail.

A raw camera that is capable of capturing a usable 12 stop linear DR range records a file that is equivalent to 0-288 IRE.

When you grade, you are going to conform your finished product to a display gamma standard anyway because that is what you are seeing on your monitor. But the range of creative choice in how you get there is far greater with the raw file than with the camera engineer's predefined compression curve. Ask any serious still photographer why they shoot raw instead of JPEG. It is the same concept. More data to work with.

3693

This is a scope shot of a log compressed video signal, Sony S-log in this case. Now it does show the camera's 12 stop response to the test chart, but 10 IRE represents the full 6dB linear DR of a photometric stop compressed by the value of the display reference gamma, for sRGB computer video 2.2, and for REC709 TV 2.4. So effectively a 10 IRE range shows a 1/2 stop linear photometric range on the monitor. You can measure this with a spot meter and a SMPTE color bar or gray scale test signal on a properly calibrated monitor if you don't believe me.

Where video camera log compression curves get compromised is at the extreme ends of the range. With the above example, you can't recover much if any visible detail for anything that falls between 0 and 10 IRE. 4 stops of the camera's linear sensor response have been compressed into less than 1/2 stop. Deep shadow detail winds up crushed to black, or if the contrast is low, muddy gray without enough MTF for detail separation. You have no control over this if you record the camera engineer's baked in log compressed "film" LUT in Prores.

With raw, if you develop the skills to use it, you have far more control over precisely where to place black clip and white clip in your image and how you render gray scale and color values within the more limited display range. You have full access to the linear capture range of the sensor to create the final image and you can shape the response curve any way you see fit to achieve a desired end result.

rze
07-12-2013, 08:26 PM
There are situations where RAW will be better, but they are few and far between and require more work. What you describe is a situation where fine detail is expected outside a normal range of exposure for important objects. You'd have to be stressing either the DR or sensitivity of the camera which won't happen on a normally produced shot. If in your customization of your gamma curve, you lift low darks, you still have the noise floor to contend with. You'd be better off giving the sensor a fair exposure range with more light than relying on RAW and noise reduction. Plus values at 10 IRE are usually crushed not lifted, we aren't talking about normal transformations.

Also I was under the impression that monitors apply gamma correction for proper viewing for the desired gamma space so the information encoded info is not actually effected, its just remapped for display. The luminance values would look the same as pre-debayered raw. RAW would only be non destructive in the luminance transformation to move darks out of extreme lows as long as the effected pixels were above the noise floor which is usually not the case.

Cinemon
07-12-2013, 09:39 PM
Somebody who grades my DNG's? :P

morgan_moore
07-13-2013, 04:28 AM
I cant download them.

Cinemon
07-13-2013, 06:40 AM
Here is another download link :

http://www.sendspace.com/file/8f6s6n

morgan_moore
07-13-2013, 06:56 AM
Sorry its the RAR I cannot unstuff!

mac

Cinemon
07-13-2013, 07:03 AM
Here are the DNG's only :

http://www.sendspace.com/file/qxe5t1
http://www.sendspace.com/file/dna242
http://www.sendspace.com/file/i749e7
http://www.sendspace.com/file/288ht5
http://www.sendspace.com/file/yrhxfx

morgan_moore
07-13-2013, 07:20 AM
I used Capture One to evaluate DNGs which IMO is the 'purest' viewer and gives you a concept what you shot.

It seems that the EXT needs a four stop boost to read at all, so basically Id say that is underexposed, but usable in an emergency

The other frames have most of the Data in the bottom 1/3 of the histo but basically have enough info to grade in any direction you desire.

Sliding the temp and tint sliders in the other four images will really go anywhere!

--
I guess you exposed to protect the highlights, fine in the normal scene, but not fine in the night ext where the point sources are many stops brighter than the ambient.

S

RobertJ
07-13-2013, 07:28 AM
Here are the DNG's only :

http://www.sendspace.com/file/qxe5t1
http://www.sendspace.com/file/dna242
http://www.sendspace.com/file/i749e7
http://www.sendspace.com/file/288ht5
http://www.sendspace.com/file/yrhxfx

Aren't those from this thread/poster: http://bmcuser.com/showthread.php?4611-BlackMagic-Cinema-Camera-Test-Footage-quot-The-Adventures-of-Molotov-Monty-quot

And besides, there's plenty of examples from Kholi on different ways to grade them.

morgan_moore
07-13-2013, 07:33 AM
Here is a screengrab from C1, the first image is the 'as shot' the following are some 'grade interpretations' you have soooo much space here. the last frame I just tried to get the skin to look like skin using the temp and tint sliders

3699

StarCoreFilms
07-13-2013, 07:33 AM
Somebody who grades my DNG's? :P


3704370537063707

Cinemon
07-13-2013, 08:16 AM
Aren't those from this thread/poster: http://bmcuser.com/showthread.php?4611-BlackMagic-Cinema-Camera-Test-Footage-quot-The-Adventures-of-Molotov-Monty-quot

And besides, there's plenty of examples from Kholi on different ways to grade them.


Oh sorry, i have uploaded the wrong dngs :(

here are the right dng's from me :

http://www.sendspace.com/file/ys8h57
http://www.sendspace.com/file/gwsvmx
http://www.sendspace.com/file/c8hj2k
http://www.sendspace.com/file/t08pxl
http://www.sendspace.com/file/p9kktg

Kholi
07-13-2013, 11:41 AM
I used Capture One to evaluate DNGs which IMO is the 'purest' viewer and gives you a concept what you shot.

It seems that the EXT needs a four stop boost to read at all, so basically Id say that is underexposed, but usable in an emergency

S

Just realized something... with the LUT I'm working on, the 4 stop increase is pretty accurate in Resolve, as well. Interesting...

razz16mm
07-13-2013, 01:51 PM
There are situations where RAW will be better, but they are few and far between and require more work. What you describe is a situation where fine detail is expected outside a normal range of exposure for important objects. You'd have to be stressing either the DR or sensitivity of the camera which won't happen on a normally produced shot. If in your customization of your gamma curve, you lift low darks, you still have the noise floor to contend with. You'd be better off giving the sensor a fair exposure range with more light than relying on RAW and noise reduction. Plus values at 10 IRE are usually crushed not lifted, we aren't talking about normal transformations.

Also I was under the impression that monitors apply gamma correction for proper viewing for the desired gamma space so the information encoded info is not actually effected, its just remapped for display. The luminance values would look the same as pre-debayered raw. RAW would only be non destructive in the luminance transformation to move darks out of extreme lows as long as the effected pixels were above the noise floor which is usually not the case.

If you are talking about normal looking digital video for the most part, I would agree. But you have the effective equivalent of a 3 stop -1, 0, +1 HDR exposure range of DR in the raw file compared to the Prores file, Log compression or not. On the raw processing side there are a wider range of possible aesthetic choices before you transcode, including HDR tone mapping, that you can't effectively do with a log file and get quality results. Most seem to want to bypass or rush through the raw processing step to get to editable video as fast as possible. This may be necessary for the time crunched "pro" markets, but it certainly isn't for a creative independent seeking to explore new aesthetic possibilities that uncompressed raw data provides access to.

The display is calibrated to the standard, or should be, but the encoded video has to fit within that display space, just as a photographer making a print has to fit what is visually important information within the limited DR and gamma of the print medium. The display provides a hard reference point , but it does not correct out of range values. Anything sent to the display that is out of a 0-110 IRE range is hard clipped at both ends. Anything in the 0-10 IRE range or the 90-100 IRE range will not hold resolved detail with good micro contrast.
In the Sony S-log example for instance do you really want to sacrifice 4 stops of shadow detail range to get an extra stop of highlight detail range? Or would you choose a different set of compromises? Once you are in an S-log encoded video file format you no longer have the choice.

The user's choice is about who determines how the range of values and color gamut in the raw image data gets mapped to the display. You or the camera maker.
Are you happy with BMC's "REC709" lut in Prores? Are you happy with BMC's film lut in Prores? If so then by all means shoot video. If not, there is an infinite range of alternate possibilities in the raw transform process, including new potential for digital cinema aesthetics that are different than either film or video can offer. It is a new world. The beauty of uncompressed open standard DNG raw is that all of the amazing image processing tools that raw still photographers enjoy are now available to film makers.

Speed and convenience have their place. All I'm saying is don't get creatively stuck in a video head space when the range of possibilities with a BMCC shooting raw is geometrically expanded. The free open source Raw Therapee image processor has 27 different tools available for mapping exposure and gray scale tonal range to an encoded RGB image for example.
The workflow I am playing with currently is transcode to TIF sequences out of RT for import into Lightworks, transcode to 8 or 10 bit video format of choice on import and do final touch up grading in LW.
I will keep the raw originals with the transform metadata attached, dump the tifs, and have the video to finish. But most of the look will be determined in the raw transform, not manipulating the video. Now this is not fast and convenient, but it is very workable and there is much potential to explore and the software is free, or low cost for LW pro license if I need the codecs.

rze
07-13-2013, 03:11 PM
Here where I disagree, if you apply the same transformations to prores and Raw, in practice not in theory, it will take more transformations than are usually necessary in the grading process to find a visible difference as long as exposure was set properly. Prores is a very robust codec that is difficult to break. Thats great that you have more parameters to adjust, but its still nothing than can't be replicated in regular transformations for baked video. What I'm saying is that RAW is a small edge, its not a worlds difference.

All the transformations to the color space or HDR effects are possible with prores it will always just be destructive to the picture, which is a loaded term especially because the degradation is very minimal. RAW will be not be destructive most of the time, but in the 0-10 IRE section it will be. Just like when the sensor clips at the same point across ISO's the noise floor remains consistent based on the amount of light hitting the sensor, so adjusting RAW exposure in post won't save you if the original source material was underneath the noise floor. In stuff that is more than 4-5 stops under middle grey at least set around 40ish IRE, that material cannot be lifted without introducing noise even with raw.

For me its not necessarily speed and convenience, its exposing the sensor properly. When you do you don't need help from raw.