I used to drop LEVELS 16-235 on the end of the FX chain to reel everything in.... but I just thought, maybe STARTING with levels before grading would alter the way the grading adjusts things? Then maybe a levels on the end to cap things? Any benefit to PRE leveling?
The answer to the question is a little unclear to me. I've always put LEVELS as the first FX in the chain and adjust it first. I've been doing this so long I forgot where I heard LEVELS should go first; I think it was in a Spotted Eagle CC tutorial video (DVD).
So I'll ask again, is it preferred practice to put LEVELS first in the chain and adjust it first?
There are clear rules for that, if you have a look to the approach that colorists tend to apply:
- first the luma correction (what can be done with levels or other tools)
- second the color correction/the grading
Sometimes you can do that in one filter - the color correction in Vegas allows also some corrections of gamma or gain. But also here - correct first the luma and then the colors.
The reason for that is that a wrong luma is perceived my stronger by our eyes compared with a wrong color setting. So to adjust the luminanc is a precondition to get the colors right.
Note that luma correction is a separate topic from using levels to conform to various video standards.
The thing is, if you do levels first to conform 16-235 then add other effects you may end up going outside the 16-235 range again. So, to be complete, you'll probably want to use something like color curves first to adjust luma, then color corrector, then levels to make sure you're in compliance. Keep in mind that the final levels correction can adjust your luma and perceived color!
The nice thing about Vegas is that you can add all three effects and then tweak them over and over again while watching the result in real time until you get it right. You don't have to "guess" the perfect luma setting first and hope it will still be ok after the other effects are added. You can go back and change it as often as you need it.
In my view, the answer is simple if you're working in an 8 bit project - ALWAYS put the 16-235 level correction at the end of the chain (my preference is on the output video bus). To see why, simply put two Levels plugs back to back in a chain. Adjust the first to drastically reduce contrast and limit the waveform to, say, 40-60. Then try to undo that and bring back full range with the second Levels plug - can't be done. Color correction plugins will be doing the same type mathematical level manipulations as Levels, although on a channel by channel basis. Why have less numbers to manipulate by limiting them up front?
Working in 32 bit gives you more leeway to "undo" a Levels correction along the signal path - but why undo it? Put it last.
Even using Wolfgang's "rule" of luminance correction first, go ahead and correct the levels first, before you begin color correction - but put it at the end of the signal chain before you begin adding CC plugins. Just my strong preference.
my question arises from my poor photography skills... a clip I shot was pretty dark, murky black shadows and all that, and even with levels at the end, it seemed too dark, I thought maybe if I put some levels at the front, it would lift the blacks up to the dark grey (contrast with the deep black letter boxing) and look better. I can bring it up with BRIGHT or GAMMA as well, was just curious about using levels for that. sounds like the wrong thing to levels crunch, then color, then levels for final.
@wwjd,
I find myself in that situation sometimes too, and I recommend using Color Curves for that - pre 16-235 Level adjustment. Making an audio analogy, Color Curves can be surgical - like a variable Q EQ. Levels is more like a hi or low shelf - although you're not necessarily removing the hi/lo luminance info, just cramming it into a smaller mathematical space.
Working in 8 bit you can still paint yourself into a corner by too much level stretching with any plugin, but ColorCurves can limit the extent of the "damage" if you watch your waveform monitor for signs of banding - when you start seeing the waveform turning into "lines."
Thanks guys. I have messed with curves a bit and understand the surgical uses, but I never seem to use it. I will mess around more with curves for that :)
Best way to learn curves and color gamma is to start with a grayscale on the timeline.
And learn to identify the tipping point for each color channel on your critically-calibrated monitor.
Only the godlike will truly get it after less than a couple of years of practice.
I'm the kid who fried a couple of CRTs trying to teach myself in the seventies.
I tend to use color curves for all three things: luma, color correction, and conformance. I'll do a general brightness adjustment with the master channel first, then tweak the individual RGB channels to get good color balance, then a few last tweaks to get the overall balance perfect again. Fortunately i don't have any clients or outputs that require strict adherence to 16..235*, but if i did i'd probably add levels at the end as well.
*When, oh when, will this 16..235 silliness go away completely and for good? There's simply no need for it in a digital->digital path. It's just a useless holdover from analog days.
I use the transcoder in Vegasaur to include the FX "computer RGB to studio RGB" as THE very last step in render to media that needs it. This way I don't have to remember it if it is in my project file for the transcoder.
I do everything (levels/brightness/contrast/colour adjustment) in a single Color Curves FX. It's fiddly but gets easier with practice. Very occasionally I need to do something (saturation, masking) in the Secondary Color Corrector.
If one corrects colors first and luma/contrast later, the color correction may not be even tend to be wrong but could also be too strong. Especially for 8bit it is a good idea to try to limit corrections to a minimum - to avoid that we bring in to much noise. That is why the trend goes more and more to 10bit even for prosumer units like the GH4 or the A7s together with units like the Shogun. Sure, one can iterate back and foreward - but as long as one shoots with 8bit the corrections should not be too strong anyway. So here the sequence of the filter is less important.
For 16..235 one will have to stick mainly for broadcast reasons. If that is not a hard constrain, one could think about to use the superwhite range too.
True, for a 32bit floating point setting we have better capabilities to recalculate corrections. But that is important for 10 or 12 bit, and a true 10bit workflow is possible in Vegas, but only using some defined codecs only, especially for uhd.
It takes a bit of a math head to do it in one filter layer, but that is preferred because there are no offsetting corrections, which could potentially cause additional banding.