Andrew Somers
Forum Replies Created
-
The ones in the hardware store have pretty poor CRI – they may be daylight balanced but still have huge spikes at certain wavelengths. For photography use, you want a CRI of no less than 90.
Try these:
-
The “real” way to make an effect as if you are looking though glass or crystal is to create it as a 3D object and use raytracing to render it.
But there are ways to cheat the effect using reflection maps and various warps or 3D layer positioning with masks.
But kind of hard to guide without a reference – do you have an image you are trying to emulate?
-
DEVIN: “I really just wanted the red sweater and the picture frames to stand out, which is why I kept it so underexposed. I should have just done the underexposing in post,”
When you are capturing images that will require later work such as compositing, you want to capture with an eye toward maintaining detail, and minimizing noise.
With that in mind, expose your images so that you are using the available “image space” to best effect, meaning using the available latitude of the media. Increase exposure just below where you wold clip highlights (clipping will lose detail), and in some cases, reduce contrast so that you improve detail in the darker areas.
Ideally, you want your histogram “filled” from left to right with image data, but with no image data being clipped of the left or right side. This will give you the greatest flexibility in post.
Caveat: When using 8 bit media, pushing this idea too far can result in other artifacts such as banding and related delta-E errors.
-
Don’t use the “color key” plug in.
Use Keylight – it’s included in AE. I have a quick sample below, and a further discussion of how to improve the key. But first:
WHY USE GREEN?
I want to say for the record that green screen is not a cure-all for key and matte purposes. As far as I can tell in this present shot, there is no hair or semi transparent objects that would benefit from a green screen key, and the objects that you are keying (the frames) are so small, that you can easily do a simple animated mask as the foreground object passes each one.Unless the camera is locked, you need to do a planar track of each picture anyway, and except for the foreground object passing in front there is no benefit to using green, since you might as well just do a corner pin for each. No keying needed.
In otherwords: it may have been more trouble to put green in the frames than you would save through keying instead of masking. (Judging by your simple foreground element of a person in a red hoodie).
Simple corner-pin-type tracked layers, with traveling masks to key the foreground red coat will give you the cleanest result for this shot, as it is presently shot (due largely to noise and lack of color channel separate, which I discuss below).
LETS TALK ABOUT ALTERNATIVES TO GREEN:
The noise in this image looks like you shot on a camera with a high level of compression. The greater the noise/compression artifacts, the more challenging it will be to make a green key. So sometimes it is worth considering alternatives to green. (If I could get a sample of your camera original, I could provide better analysis).
Green is not the only valid choice here. Flocked black could work in some cases, white can work in some cases.
Flocked black (dense velvet or duv that reflects no light) are good if you have glass on top and you want to retain the reflections. You can cornerpin/mask and ADD the images that go into the frames, and the reflections remain naturally. You may need to use traveling masks to matte the foreground element, though in some cases you can use a Luma key with a combination of garbage masks – again, depending on the shot.
If instead of reflections you want to retain shadows over the images from other object in the scene, then you could use WHITE screen instead of flock black, and then you would use MULTIPLY to place the images in the frames, and the shadows and relative scene lighting on the white screen would be retained.
Now, there are certainly plenty of good reasons to use green – especially if you have a foreground object with complex edges (i.e. hair) that is hard to manually mask.
But if you choose green, there are parameters that you need to stay within for successful keying.
NOW LET’S DISCUSS YOUR SHOT, AS SHOT:
It is not that your green screens are too dark per se – it is a common misconception that a green screen must be “bright”. Actually, a green screen can be two stops UNDER meter for a scene that is being exposed at meter (i.e. green screen can be in zone 3 – so if you spot meter an 18% grey card and expose it in zone 5, then spot meter on the green screen can be two stops below).This also assumes, however, that your camera and the compression used on the footage, is noise free enough (it certainly isn’t in the example shot you posted).
As a general rule, if you want your scene dark, you should still shoot it brighter (so long as you don’t blow highlights), and then bring it down in post. This will help to eliminate noise issues and maintain detail in the darker areas. This is particularly important when shooting on dSLRs and cheaper camcorders that record to a highly compressed 8 bit file.
The biggest problem you are facing here with your shot is the extreme noise in the green screens due to the low exposure level.
Incidentally, it is a mistake to attempt to color correct a plate before keying – i.e. do NOT put levels or curves in before Keylight. In most cases, Keylight (pr primatte) should be the first plug-in.
SEPARATION:
The other important aspect is the separation of green from the red and blue channels. This is why proper narrowband green needs to be used in the screen material. You can’t just stick any old green paper or whatnot in for a green screen and expect good results.The green paper you used isn’t all that green. If you examine it in linear colorspace, you will see that the GREEN channel is only about one stop brighter than the RED and BLUE channel.
See that your green channel is less than twice as bright as the other two, i.e. the green is less than a stop brighter than red and green.
Here are some real world examples of want to see in a green screen. In the first two cases, the green screen is exposed at meter (i.e. a spot meter reading of the screen is equal to a spot reading of an 18% grey card). In the third example, the screen is exposed 1 stop under meter, and in the fourth, it is 2 stops under. (Note that as a general rule for practical reasons (such as noise), you typically don’t want the darkest useable portion of the screen much darker than -2 stops. This can be important when there are cast shadows you want to preserve and avoid noise.)
In these examples, the green channel is three to 4 stops brighter than the next brightest channel, and more than 6 stops brighter than the darkest channel.
But in your case, your green paper element is not narrow band, with less than a stop of separation. If you are looking for a paper product, try something that is “florescent green” as a way to get a very narrow band of light reflection.
TIP: You can test for the bandwidth of a particular color by taking a full screen image of it with a dSLR and then looking at the RGB histogram in the camera. You want to see both the red and blue spikes far to the left of the green spike, with as much distance between the green, and the red/blue as possible. Take your camera to the art store when choosing the color, and check with the camera (place the paper sample under a light source type similar to your shooting setup).
Thus with your shot, exposing it brighter (to reduce noise), and using a more narrowband green element, should solve your issues.
Regardless, don’t use the “color key” plug in. Use Keylight or Primatte:
Here’s a rough key after spending about a minute using Keylight:
Now, if this were on your QT file, and not a still, you would probably see crawling noise at the edges of the picture frames due to the noise in the element.
If I received this “as a shot to do”, I would probably not use the green for a key at all, and just comp over the green with the images, then use animated masks for the foreground person passing by.
-
Olson: “You should have your edit locked before you start your VFX work”
If we waited till lock, the feature would never get done, LOL. 🙂
-
Odd – no such rule really exists.
I know VFX shots that are *minutes* in length that no one would know was a VFX. I know other VFX shots that are less than a second that are completely lame.
It depends on context, quality, and a bazillion other factors. Forget such a guideline, it is meaningless.
-
Welcome to the Nikon family 🙂
Yes I agree with not getting the 18-105 kit lens, it’s not one I recommend.
Lenses are something I usually buy used.
The 18-70mm f 3.5-4.5 DX AF-S is surprisingly good, and available used for $150. It’s better than typical “kit” lenses, but surprisingly sharp with decent bokeh.
But to really take full advantage of the D7000, get the good glass, which generally means zooms with a constant f2.8, and primes that are f 1.4 (for 85mm and wider, or f2 to f2.8 forlonger primes).
The 85mm f 1.8 AF-D is also pretty good, and relatively inexpensive used compared to the 85mm 1.4.
The 50mm 1.8 AF-D is sharp and cheap, BUT the bokeh is terrible, the 1.4 is better. Allegedly the newer 50mm 1.8 AF-S has better bokeh but I have not tried it.
I have the Sigma 10-20mm ultra wide zoom, and I do like it, though it’s a bit slow – you might prefer the Tokina (at twice the price).
The *excellent* tele-zoom of course is the 70-200mm 2.8 VR II (at $2500). But you will be surprised at the images from the 70-300mm f 4-5.6 VR. While not as fast as the 70-200, it is also far lighter and easier to carry/handle. The VR version of the 70-300 takes good images with good bokeh- but do note that the NON-VR 70-300 versions are pretty terrible with awful bokeh, so only this version of the 70-300mm is acceptable IMO:
https://www.amazon.com/Nikon-70-300mm-4-5-5-6G-Digital-Cameras/dp/B000HJPK2C
You can also get the older 80-200mm 2.8 AF-D used for around $500, though as with the more modern 70-200 2.8 VRII, it is a huge heavy beast. I find I use the 70-300 f4 more often.
If you want a lens that will make you hate almost every other lens you’ve ever used try the nikkor DC series – the 105mm and the 135mm DC are so unbelievable that you’ll never want to use a kit level lens again:
The 85mm 1.4 is also in this class of “amazing” (and you pay for it, LOL).
-
I *believe* that a recording must stop before switching cards.
As for new D7000s, It is availabel new with the kit lens here:
And Amazon says the body only will be in stock in two days here:
Photozone has the body only for $200 MORE than list price, LOL:
IMO order the Amazon one and wait a couple more days.
WhenI originally bought mine I waited over 6 weeks… 🙂
-
The D5100 and D7000 are very similar in terms of image performance.
What he D5100 does not have (and these things may not matter to you):
1) 5100 is mainly plastic, the D7000 has a magnesium alloy frame (though to be fair, the mirror box is still engineering plastic on the D7000).
2) The D7000 is weather sealed.
3) The D7000 has a built in autofocus motor, so it can autofocus many Nikkor lenses that the D5100 cannot. The 105mm DC and 135mm DC are two I use and need the AF motor for, along with my 80-200 f2.8 AF-D.
4) The D7000 has dual UHX SD card slots – a UHS card transfers data much faster than a regular SD card.
5) For stills, the D7000 has a faster shutter (1/8000 vs 1/4000), better bracketing options (important for HDR), and a far better AF system (39 points vs 11).
6) The D5100 uses a different battery, the EL14 which holds about HALF the power of the EL15 that the D7000 uses. Since you mentioned long periods on a charge, this may be important to you.
You can compare nikon cameras here:
https://www.nikonusa.com/Nikon-Products/Digital-SLR-Cameras/index.page
-
The additive layer needs to be above the normal layer.
“wider choke” means less shrinking.
