REDSPACE vs REDlog or PDlog

[Levan Bakhia]

So, what is better? I read a lot from read team that it is always better to use redspace, but what I see in practice is that redspace is very good if just to get the normal look, but if you need to make more adjustments, I find it is better to just do some corrections in REDlog and output it as dpx files and then do secondary correction to it. While with redspace it seems that you are more limited. Redspace looks better right away, but seems like a bit less details at the end. so help me here, am I wrong ?

 

[Gunleik Groven]

you must differensiate on set and in post...


in othert words: what you see sounds right

g

 

[Mark Toia]

What we have found if you just work with the 2k Quicktimes only instead of rendering out DPX or tiffs is.....

RED SPACE has a great look straight up, but I'm finding I have to grade backwards if I want to get more info out of the blacks. RED SPACE is sort of grading it nicely for you in camera. As if the curves have been tweaked a bit. Looks good on the onset monitors but a little controlling when you want to grade that 2k QT later. Client has already warmed to it, next thing your stuck with the REDSPACE look. Makes it hard sometimes.

REC709, is a little flatter in color and contrast. It's a good middle ground. Clients ask if it's going to be graded, you have a chance to say yes.
Even though I like looking at REDSPACE images on set. REC 709 is i think is better to work with if your using the 2k QT's only and not outputting to DPX or tiff going though Scratch etc.

RED LOG looks like flat crap on set, flat crap in the edit suite. I use it only to see what range I have to play with in a contrast scene. I Grade in REDLOG via RED ALERT if I go to DPX or tiff output. Which is not often at all, only if I have under exposed a shot and I want to see information in the blacks. LOG you seem to see everything.

There are going to be many answers for you on this forum, Not saying my way is the best way, it's just one of many ways.

in Short, REDSPACE looks great on the little LCD screens and onset monitors.

REC 709 gives you a more neutral starting point if you going to grade in FCP and Colour later.

REDLOG is for the time you think you have something to save.

No downsides to be honest.

Hope this helps out a little bit.

 

[Esteban Sosnitsky]

As Gunleik stated, it depends on the project and what you are doing. My best guess after all is to experiment and see what suits you best according to the technical knowledge as a base. Also I think one should get the info from the sensor in red alert as expandable as possible and finalize color correction and look on a different software like After Effects, Color, etc.

 

[Irri]

I’m really looking forward to the white paper on this because it is hard to find a clear explanation of them all. Here’s my best guess:

REC709:
Assuming you are going to HD then this should look correct as HDTV is REC709. If you transfer to something like a 10bit DPX then you have to discard a lot of information and you may be better off with:

REDLOG:
This is a container format. It is used to retain detail in the lower registers by discarding more in the highlights where there is an abundance of information. It should be used with LUT so that it looks like REC709 (i.e correct), but many people seem to use it without one.

REDSPACE:
RED images can look a little desaturated and low contrast before they are graded. REDSPACE gives the image a sort of quick grade to make it look a little like REC709 with a grade.

 

[Christian Edwards]

http://www.reduser.net/forum/showthread.php?t=425&page=7

we need a Red Color Consortium

 

[SF Geek]

You better protect for your highlights if you're shooting redspace to 2k quicktime output. It seems that there is a more filmlike curve applied to the high end which causes clipping to occur sooner but more smoothly. If you're not doing any color correction in redcine or red alert, then you're not going to be able to get that information back.

 

[Fredrik Callinggard]

If you off-line and then on-line for a final grading you should use REDLog, so you have the most detail from your project for your grading session.

For anything else it's up to you. I can't help you there since my only path with productions are with a final grading at the end, but REDSpace seems to be a good container.

Fredrik Callinggard

 

[Dan Hudgins]

REDLog

REDLog

If you off-line and then on-line for a final grading you should use REDLog, so you have the most detail from your project for your grading session.
Fredrik Callinggard

I am working on a uncompressed 16bit (48bit) TIFF import into my "freeish" 4K (0K to 15K+) DI post system, and seem to like the REDLog TIFF best so far, good for low noise in shadow areas when adjusted to video gamma 2.4.

 

[Irri]

Shouldn't a 16bit tiff be able to hold all the information needed and thus negate the need for a logarithmic conversion?

 

[Irri]

You better protect for your highlights if you're shooting redspace to 2k quicktime output. It seems that there is a more filmlike curve applied to the high end which causes clipping to occur sooner but more smoothly. If you're not doing any color correction in redcine or red alert, then you're not going to be able to get that information back.

Yeah, redspace (right) seems to go earlier without tweaking:

 

[Christian Edwards]

So, what is better? I read a lot from read team that it is always better to use redspace, but what I see in practice is that redspace is very good if just to get the normal look, but if you need to make more adjustments, I find it is better to just do some corrections in REDlog and output it as dpx files and then do secondary correction to it. While with redspace it seems that you are more limited. Redspace looks better right away, but seems like a bit less details at the end. so help me here, am I wrong ?

man check Red center: red centre #008 great explanation

http://www.fxguide.com/redcentre

 

[Dan Hudgins]

de-Bayer sub values?

de-Bayer sub values?

Shouldn't a 16bit tiff be able to hold all the information needed and thus negate the need for a logarithmic conversion?

Log may be better than linear if the de-Bayer is making sub-values that are more than 16 bits, that is if they are using real or 32bit words for the intermediate values for the pixels, you can end up with more than 12 bits of data through interpolation.

If you downres 4K to 2K you should get data going from 12bits to 14bits per pixel (not fast sub-sample, but sum of four pixels), with the de-Bayer taking data from maybe 16 pixels for the chroma around the output pixel you might get "data" at more than 16 bits, so if you pull the dark areas up, you might see a little less noise (step size) in the shadow areas with log 16bits over linear 16 bits?

 

[Irri]

Ah, I think I see what you mean. 4K to 2K down sample gives us 4 times the information, so 12bit becomes 14bit. We also get more surrounding pixels in the debayer as our single pixel is made up from 4 pixels?

 

[Emery Wells]

Shouldn't a 16bit tiff be able to hold all the information needed and thus negate the need for a logarithmic conversion?

Indeed

 

[Brook Willard]

I'd love a white paper.

Linear = 0 - 4095 in a linear sequence.
REDlog = 0- 1024 in a logarithmic sequence. I have a value transfer sheet for REDlog.
REC709 = a standard color space... shouldn't be hard to find a proper white paper on it.
REDspace = ???
PDlog = ???
Camera RGB = ???

I'd love an excel sheet with 0-4095 linear in the left column and every other color space RED supports in columns to the right.

 

[Irri]

Indeed

Is that "indeed" as in "technically it should" or as in "sigh, you'd think so, wouldn't you?"

REDlog = 0- 1024 in a logarithmic sequence.

Even outputing to a container with more than 10bits?

 

[Brook Willard]

To my knowledge, yes. The sheet I have only lists values from 0-1024... not 4095.

As for why REDlog is made up of 1025 values instead of 1024... I dunno.

 

[Dan Hudgins]

Should if it is done right?

Should if it is done right?

Ah, I think I see what you mean. 4K to 2K down sample gives us 4 times the information, so 12bit becomes 14bit. We also get more surrounding pixels in the debayer as our single pixel is made up from 4 pixels?

The sensor makes 12 bits, but those get converted into 48bits.

How well those 48bits get filled up depends on the way the software works for the de-Bayer. You can look at the word data in the TIFF file to see if the lower bits change smoothly or in jumps to see how much of the summed data is still there, or if it was rounded off to some extent.

If you have some sample 4K and 2K TIFF files made with the RED software directly from the RED CODE I can look at the data and say more about how the data looks. One can see if the lower bits are set to all zeros and stuff.

 

[Dan Hudgins]

Histogram gaps?

Histogram gaps?

To my knowledge, yes. The sheet I have only lists values from 0-1024... not 4095.

As for why REDlog is made up of 1025 values instead of 1024... I dunno.

You mean 0 to 1023 only?

I am still working on my TIFF import, but from what I can see in the REDLog TIFF file I have use higher bits, the values look higher than 1024 in the data since I read 22000 to 45000 range or so.

Do you have a sample REDLog TIFF file direct from the RED software I can look at with a wide range subject to see the bits used?

One needs more than 10 bits for images that will undergo a large contrast boost to avoid tone banding. It is best to save all 16 bits from the de-Bayer in the TIFF file to hold the gradations added during the pixel interpolation.

Once the images are color corrected, you still should use more than 10 bits since the film recorder will do some gamma and curve adjustments to compensate for the CRT and film used. To end up with 10 bits per pixel on the film print you probably need about 12 input into the film recorder maybe, unless the data is uncompressed and dithered and 4K, which would allow some mixing of the pixels in the film recorder lens and film to make in between tones.

Whenever you bost a part of the brightness curve you make gaps in the histogram at some bit level, and if you go from one system to another and make adjustments you make more gaps, so it is best to hold as many bits as you can.

Outputting 4K in the film recorder should give a two bit advantage over outputting 2K with the same 10 bit D2A converter since the four pixel groups will mix in the film and film recorder lens to make in between tones.