Photography tutorial for textures?
Hi I can find alot of tutorials for hand painting textures and alot for photography basics but i'm drawing blanks for tutorials specifically for how to take photos for video games and film textures.
Basically i'm getting told by different people to not use a flash because we can't have reflections and shadows and things like that but then others say use an off camera flash otherwise stuff won't be in focus or very detailed.
I'm finding lists that say requirements for texture files and photos but no idea if they are right or not such as this.
http://photography.tutsplus.com/tutorials/12-tips-tricks-for-shooting-great-textures-and-a-free-texture-pack--photo-5373
Is this 100% right or what?
Above could mean either since the 60 being a higher number like 120 is above 60 but 1/30 is a longer time than 1/60 so is above in length of time.
Basically i'm getting told by different people to not use a flash because we can't have reflections and shadows and things like that but then others say use an off camera flash otherwise stuff won't be in focus or very detailed.
I'm finding lists that say requirements for texture files and photos but no idea if they are right or not such as this.
http://photography.tutsplus.com/tutorials/12-tips-tricks-for-shooting-great-textures-and-a-free-texture-pack--photo-5373
Is this 100% right or what?
They say this but whats 1/60th and above mean? does it mean the 60 should be a higher number like 1/120 or the 1/4000 my new camera can do or do they mean the 60 should be a lower number like 1/30 or 1/15 or something?try to use shutter speeds that start from 1/60th of a second and above
Above could mean either since the 60 being a higher number like 120 is above 60 but 1/30 is a longer time than 1/60 so is above in length of time.
Replies
http://www.polycount.com/forum/showthread.php?t=152125
I'm working on something more in depth but I don't know if there's a lot of demand for it.
And I wouldn't use off camera flash unless you really know what you're doing. Though I can't see why you'd need to if you're using a tripod.
Very awesome
Either way thanks alot
1. Shoot in cloudy or hazy conditions without direct light
3. Don't buy/use wide optic or much zoom. Something in the middle does crisper shots without corner blur and vignetting.
Now you could just auto stitch multiple shots instead of using wide angle .
4. If you shoot ground wear white snickers, they wouldn't let your camera overexpose and could be easily auto masked to remove . X-rite passport or white/grey/black card still may be helpful too
5. Any camera is acceptable including ones in nice cell phones. Fixed lenses are better then zoom ones.
One critical thing is shot to shot speed although, cheap soap boxes often do 5-10 shots then freezes.
6. GPS/GLONAS is a nice and useful thing to work with texture photos. Helps to keep series auto sorted.
7. Don't use RAW and super huge megapixel count. You would rather waste your time and hard drive space.
Can you provide some example shots please? Because this doesn't sound like good advice if you want to get high quality textures.
A random one from my phone:
https://www.dropbox.com/sc/gf4330xsosd47fu/AAAy2JDS2likF3_vWyygTxd3a
I just think a good texture is not one where you could distinguish every microbe and repeating in every 2 meters but rather a series of photos covering an extent of 10-15 meters and allowing to extract depth info nicely
https://www.dropbox.com/sc/qpd2m35lik2brz0/AADiICIRheta0wCUoUujOd2za
For that specific task too hi res + RAW would make you a lot of extra pain
ps. I personally never use all those textures people do available in texture stocks. I am an environment artist and what would I do with a rock or ground picture 2x2 meters. Tile it like a crazy? They are mostly useless
The example texture above suffers from over-sharpening (it makes it look more detailed but it isn't real resolution, also halo artifacts), poor dynamic range (blown highlights), image quality that falls off towards the edges and jpeg compression artifacts. All of which would be improved by shooting with a high quality camera and lens, in raw, and carefully processing (which is actually very fast in lightroom, process one image in ~1 minute and copy changes to all other images in set).
A high megapixel camera (16mp or more) is also a good idea. The reason for this is how the anti-aliasing filter and RGB arrays in digital sensors work. Firstly, each pixel in a camera only records one color (red, green or blue) so each pixel can not record the full photon information, because of this, you need groups of pixels to accurately represent the scene. If you look closely at gnoop's texture, you'll notice that most detail is in solid blocks of 4x4 pixels, meaning the resolution is roughly 1/4th that of the pixel size, this is due to poor resolution of the camera, or jpeg compression artifacts, or more likely, both.
Anti-aliasing filters further blur the image detail to avoid moire artifacts. All this means that a 16MP image rarely contains 16MP worth of detail, more realistically it's like 8-10MP of detail but only in ideal situations (shot at base ISO, at the lens' best aperture, with no motion blur). A 4K texture requires 16MP of detail, so even a 16MP camera is light on information for that purpose, a 24MP or 36MP sensor would be better. 8-16MP with a good lens is fine for 2K textures.
While some of the better cell phone cameras are perfectly fine for general use and posting on facebook, if you wan't high quality, high res textures, I would invest in a dedicated camera.
The quality and resolution of modern cameras is way above the texture needs.
And in all scenarios you would need more close shots in a series with subtle parallax and not so hi resolution than a fewer extremely hi res ones.
RAW definitely has advantages but in real production where time is matter you would just waste your time to keep, adjust and export from them.
It's just from my personal experience. Because even 8 mpix camera does much, much more dense pixel per meter coverage than actual textures do.
And when you scale to real texture you just suddenly notice that all that sharpening and halos make no difference at all. So why bother.
The only thing where they could make problems is ads, logos etc. But most probably you would still want to use something like that http://www.imagaro.com/
ps. Sometimes I see Mari examples with super hi res textures. Have no idea why they do them, since even in movies its a waste of time mostly. But who knows, perhaps it's where you need big camera
Test cameras:
Olympus EM5II with Panasonic 25mm 1.4
Sony RX100 III
Samsung Galaxy S4 (some phones have better cameras than this, but it's a very common phone and also what I happen to own)
Here are 100% crops at 4K, top: Olympus, mid: Sony, bottom: Samsung:
100% at 2K:
100% at 1K:
Fill size shots:
https://dl.dropboxusercontent.com/u/499159/restestoly.png
https://dl.dropboxusercontent.com/u/499159/restestsony.png
https://dl.dropboxusercontent.com/u/499159/restestsam.png
Of note, the 16MP EM5 II delivers better detail than the 20MP Sony RX100 III despite the Sony having more pixels. This is because the EM5 has better (larger) pixels and a better lens. Both, to my eyes at least, look noticeably better than the phone at 2K and 4K res, at 1K, its hard to tell them apart, but it's worth remembering that 1K is 1/16th the size of 4K. It's worth paying attention to not only the sharpness in the Samsung shot, but also the color and tonality - brighter elements saturate the sensor (overexposure) much easier than the larger sensor cameras.
My final recommendation: If you don't need large textures, or if a phone is all you have on you and you see something cool, go for it! If you're going out specifically to take reference photos, you might as well bring a proper camera. The RX100 is small enough that you can put it in your pocket and take it everywhere. A compact M43rds system isn't hard to put together either (EM10 + Panasonic 20/1.7 pancake?).
RAW v JPEG
Really, the question here isn't even raw vs jpeg, its processed vs unprocessed images. If you're taking texture reference, I assume you're at minimum adjusting white balance and exposure (on a calibrated monitor) when you get back to your workstation. If that is the case, working from a raw or a jpeg file doesn't affect workflow speed in any way other than the time it takes to import and save files. If you're processing, working from raw simply means working from a higher quality source, which is image processing 101 level stuff. If you're going to manipulate content, you shouldn't work with lossy, compressed file formats.
If you start with a jpeg, on capture, the jpeg is compressed. If you save as jpeg after making white balance/exposure adjustments, you're compressing a 2nd time, and by time you get it into game, you've compressed the texture at minimum 3 times, which is a bad idea for obvious reasons.
Basically, when shooting texture reference, the majority of your time goes into traveling to the location, taking the photos, and cataloging them. Editing from raw or jpeg is of such little consequence that it shouldn't even enter into the equation. If your workflow requires you to spend a significant amount of time on each shot, you should look into different software/workflows. In lightroom, processing reference photos is as simple as doing the first of a set, hitting copy, selecting all in the set, and hitting paste (this applies to both RAW and JPEG source workflows).
It's good enough
This is where it gets really subjective. What is good enough will vary depending on the project, artist, intended use, etc. What is good enough is also a moving target. Not that long ago using 1K textures seamed absurd. In a few years time, as GPUs get even more VRAM, and 4K displays are more common, higher resolution textures will be in demand. Being able to reuse your texture library later will probably save you a lot more time than trivial differences like working with raw or jpeg.
Personally, I think its a good practice to work with content as high quality as possible without significantly delaying production. To me, using a different camera, or shooting in RAW v JPEG falls into this category. If you're putting in the effort to make content, why not make it as good as possible? These days we're lucky, rather than lug around a huge DSLR, we can take a mirrorless or a high end compact point and shoot which deliver extremely good quality.
Additionally, there may be other, different uses for the content in the future that you're not able to predict, so having high res, high quality source rather than merely good enough for today is a safe way to hedge your bets.
Because making a texture with double the resolution doesn't take twice the time. Nowhere near it in fact. There isn't a linear boost in production by working at the bare minimum res. On the inverse, when working in film, you don't always know beforehand if the asset you're creating is going to be shown in a closeup (this can be said for certain types of games as well).
Excuse my newbiness, but do I still need to adjust this manually ? If so how often / what situations should this be done in ?
Auto white balance is too inaccurate. Get a white balance card or an x-rite color passport. White balance cards are relatively cheap and you only really need 1 shot per session (or when the lighting conditions change) so it's not like it takes much effort.
If using JPEG, you can use the custom white balance setting to sample the gray card for each set. Not all cameras have this, but I think the 450D does.
A grey card is good, but a full color checker like the x-rite color passport recommended above is a better idea. The reason for this is exposure calibration.
Camera's autoexposure function works by trying to make sure the average of the frame is 50% gray. This works for general photography, but is not an accurate way to capture color information. For isntance, asphalt or other dark surfaces will be overexposed, while sand, snow, etc will be underexposued. Shooting in RAW and using a color checker with various shades of gray so you can calibrate the gray point of exposure is essential if you want your textures to have accurate color/brightness information.
Again, you can correct exposure from JPEG, but when you clip highlights or shadows in a JPEG, the information is lost forever. On the other hand, RAW files typically contain 10-14 stops of dynamic range information, which means you can generally change exposure levels up to 2-3 stops without seriously degrading image quality.
Now, usually when shooting references photos, you'll be working with overcast lighting so lighting contrast will be low, and most cameras will be able to record much of the dynamic range. However, both of the examples in this thread shot with cell phones (gnoop's and mine) have clipped highlight detail that can't be recovered. Certain subject matter, like a black asphalt road with white paint, with be difficult to capture with a cell phone/jpeg's limited range, as the auto exposure tells the camera to make the road brighter than it should be, causing the white paint to clip. You can attempt correct this in camera by shooting in manual mode or adjusting exposure compensation (even then you're eyeballing, not setting the correct exposure), but it's a lot easier and more accurate to correct via raw later.
So, how do you actually calibrate the gray point? For the purposes of white balance, it's easy. In Lightroom, simply use the white balance tool to color pick the grey point on your color card, this should neutralize the white balance. For exposure, you'll have to cross reference the sRGB values, which can be found here: https://en.wikipedia.org/wiki/ColorChecker - it's important to check multiple values as well, not just mid gray, because your camera (especially if using jpeg) probably bumps up the contrast, which will throw the values off. Once you've done this for the first shot in the set, copy your settings and paste them to the rest of the set.
For different colored objects, you'll have to re-calibrate the exposure, and the white balance too if the lighting changed.
As of 1, 2 or 4 k . it's matter not how much pixels your textures have but rather the size of texel ( in cm/inch ) you are using. Or to be more clear, what space the texture would cover.
And if you you take your phone close enough to a subject and it does speedy series you could do whatever you need.
As of RAW. I did a lot of RAW before. But gradually I became tired of keeping,sorting and managing many thousands of them. Lightroom starts to slow down with really huge base. Keywording is a pain there actually. They take more drive space after all. You can't drag and drop from Lightroom and have to export first.
And one day I though why I do all of this . My textures are shot with mild ambient lighting and there is nothing to fix with the exposure. I do sharpening and scale down a lot nevertheless.
So Last time I open Lightroom was pretty long ago . I prefer http://www.photools.com/
And one more comment about resolution. Every extra pixel of 8 gig video memory would rather go to make my textures cover more extent and be less repetitive.
And with such approach it's getting a kind of problem in Zbrush with keeping too much of unnecessary extra resolution
But I agree if you are going to special reference trip you better take normal camera. But in random occasional texture capturing where you are probably take best of your textures, any camera is acceptable.
I get what you're saying though, you don't see the benefit for your purposes, that's completely reasonable, you've worked with your system for a while and understand your own requirements. Still, for beginners, I think its important to understand how to get the best results before cutting corners. This applies to anything art related, you should understand the rules of anatomy before breaking them, you should understand how to bake a normal map properly before attempting to paint out errors, etc.
As far as the RAW stuff goes, I will reiterate once more, its not a question of RAW v JPG, its a question of processed vs unprocessed images. It sounds like you don't do any processing, which I find rather odd, but in that case, using jpeg won't make a big difference.
[ame="http://www.amazon.com/X-Rite-MSCCPP-ColorChecker-Passport/dp/B002NU5UW8"]Amazon.com : X-Rite MSCCPP ColorChecker Passport : Digital Camera Accessory Kits : Camera & Photo[/ame]
Is it:
option 1: Take a 1st photo of the chart, then take a 2nd photo of the material/object
option 2: Take only 1 photo of the material/object, but with the chart in the scene
If it's option 1: Couldn't the lighting information be different between the 2 photos? (1st photo being of the chart, and the 2nd being of the material/object)
How can you accurately adjust the 2nd photo if you don't have the chart in the image?
If it's option 2: If the chart is in the photo then it's in the way of what you are trying to photograph.
Calibrating based on the chart detailed here: http://www.polycount.com/forum/showpost.php?p=2320565&postcount=13
You could take a photo of just the chart as well, but only if your white balance and exposure is locked down, otherwise auto exposure and auto WB will mean different readings for every shot afterwards which negates the benefit of using a color chart.
If lighting changes significantly (the sun comes out, you move to a different location, etc), you should take another shot of the chart.
All of the photos are off of out camera, some materials like my white siding and my orange stained bench are really really off. Again, if you understand how a camera meters a photo, it's shooting for even gray, so the uncorrected results make sense, and are often not really an issue for regular photography (you'll probably fix the color to taste in post anyway). However, if you're trying to capture accurate albedo information for a library of textures, this sort of variance is a serious problem.
Imo should be more neutral. Would probably more or less ok with intense blue sky but under more neutral cloudy lighting it would be too warm in an engine. Especially asphalt textures.
Also that white door while having right albedo value probably in real game engine would always be blown out. Whatever adaptive tone mapping a game would use, a human eye in reality almost always see details in both blacks and white colored things. I would not follow white albedo values so strictly
These might be slightly tinted towards the yellow spectrum, and that might be due to my cheapish color card (I wanted something small I could fit in my pocket, hard to find, but it seems the values vary from gray a couple RGB values each - which makes me generally skeptical, I've another card on the way). Even still, the results are much more accurate than auto exposure and auto white balance, which was the point of the exercise. Its really impossible to look at calibrated content, without the source material in hand and decide if it's correct or not. You can tell me whether you like subjectively it or not, but there's no way you can objectively say with any accuracy whether it is actually correct.
Right = calibrated left = out of camera. You may prefer the one on the left, but that's because again, you don't have the original item to cross reference. If you actually check the values, the one on the right is shifted significantly (about 15 rgb values) to the blue spectrum. Now, I may do some further tests with another card just to make sure mine isn't totally off, if that's the case, it's simply a tool problem, not something that would invalidate the calibration workflow.
Eyeballing exposure values doesn't make any sense either. Having consistent materials values is key for art direction. White siding, paint, etc should certainly be white (about 95%) in game. If it's blowing out constantly that's a lighting/post issue. If it "looks" too bright, your monitor calibration is probably off (most monitors at full brightness are too bright to accurately display color). It's a really common mistake to make albedo textures too dark because of how they "look" to the artist in photoshop. You gotta test it out in engine, here is what it actually looks like when I construct a material with this albedo value:
No blowing out, the highlights are not clipping, no problem here even with IBL and a couple direct lights. When you decide a few random materials shouldn't have accurate values, you start to throw the whole system out of whack and may as well just eyeball whatever looks good instead.
If you rely on your camera's auto exposure, the only thing you can be certain of is that your values are consistently inconsistent and incorrect.
If you don't want to take my word for it, read what S
And I don't believe anything beyond $100 X-Rite passport (color checker). Although read somewhere that it's aging too after few years.
As of whites exposure I agree you can adjust lighting intensity but you would make everything darker too missing all that "sunny" feeling. The problem is that rendering math follows a camera approach, not a human eye perception. And all that HDR bloom only ads extra exposure, not other way around. So I still believe it's not a mistake , rather a try to match a picture with a human eye perception and not random low dynamic range camera.
At least in our own engine it clips whites a lot. U4 seems doing it right although.
In fact I don't see much value in all that precise texture calibration people look so obsessed. It makes zero difference in a game actually if you textures varying slightly from 100% precise values and you can just go around those few examples on engine sites. In real world things are varying too.
Moreover I believe a key thing is believing to your own taste and eyes once you have properly calibrated screen.
tripod?
Sometimes flash helps. If you set it to +3 then close it with a hand or a few thick a bit crumpled white paper sheets. Would still illuminate surroundings pretty well
But what should the settings be for:
1) aspect ratio: 3:2, 4:3, 16:9, or 1:1?
2) Flash Comp
3) Single shot AF, or Manual Focus
4) Focus Area Wide or Center
5) Exposure Comp
6) Iso setting
7) White Balance
8. F stop
9) Shutter Speed
Thank you.
Do some test shots of different ratios, and see which has least edge artifacts... vignette, aberration, blurring, etc.
Avoid flash unless you have something to bounce off, or diffuse it through. Makes too sharp lighting.
Auto focus is generally better. Hard to judge manual focus in those little LCDs.
Sick of phone typing, sorry.
First off, shoot in raw to make processing easier.
1) aspect ratio: 3:2, 4:3, 16:9, or 1:1?
3:2 is the native res, not much reason to do anything other than that. 3:2 gives you the most room to crop, while all other aspect ratios are pre-cropped.
2) Flash Comp
Not revenant, you shouldn't be using flash at all.
3) Single shot AF, or Manual Focus
Single AF
4) Focus Area Wide or Center
I use center with my RX100III and place the focus point where I want, this is the best way to ensure proper focus. Wide area may result in something other than your subject being in focus.
5) Exposure Comp
If you shoot in raw, leave this at 0. If you're shooting jpeg, you may want to adjust it per shot. As shown above, auto exposure can cause bright surfaces to be under exposed and dark surfaces to be over exposed, which you would adjust the exposure comp to compensate for.
6) Iso setting
As low as possible, in the 100-200 range if you can without causing motion blur.
7) White Balance
If RAW, set it to auto, doesn't really matter. If JPEG, set it to the closest setting to your lighting condition. For instance, if shooting under overcast lighting, set it to cloudy, or set it to custom based on your gray card. For JPEG, the WB information is not recorded (it is in raw) which means if you shoot with auto WB in jpeg, your color temperate will vary per shot which will make color balancing difficult.
8. F stop
Generally around F2.8-5.6. I would probably leave it F4 unless you're dealing with low light (then go to 2.8 or lower) or if your object isn't flat (or is very small) then stop down to 5.6 or even 8 to make sure the depth of field is sufficiently wide. You'll have to experiment a bit and see what works best.
9) Shutter Speed
This depends on the focal length you're shooting at. The safe range is typically 1/focal length (35mm equivalent) so at the 24mm end, 1/25 should do, while 1/80 should do at the 70mm end. If you have enough light to keep the ISO low and use faster shutter speeds than this, by all means do. You should experiment with this too, as it tends to vary a bit depending on the camera, your technique (how steady you hold it) and how effective the image stabilization is (from my experience the IS on the RX100 isn't super effective or reliable, so I wouldn't push the speeds too low). To be extra safe, double the 1/FL rule and do 1/50 and 1/180 respectively at either end of the zoom range.
This is really a backwards way to think about it. If your bloom is blowing out your bright surfaces, you should adjust the bloom, or the tone mapper, not the other way around. If you do bloom in HDR, set your bloom threshold to an HDR value (like 2 or 4) so that only things that are *really* bright, like the sun, bright reflections on metals, etc, bloom out. Bloom applied to everything that is sort-of bright looks terrible anyway.
If you go around selectively darkening textures which bloom out in bright light, you're going to throw off the values in shade. In shade, your darkened white painted texture will look gray rather than white. I've worked on projects where we do this, adjust the textures to match the lighting instead of using logical values and adjusting lighting/post to suit, and its just a really bad way to work. You're constantly chasing down textures and editing them but you're only treating the symptom of the problem, not the source. When you have logical material values, lighting is actually a much easier and more straight forward process.
I agree that surfaces in reality vary, and that correcting to perfectly accurate values can be obsessive. However, with certain surfaces here like the white siding, the value was nearly 3 stops off. This is a massive, massive difference. One stop too bright means the value is 2 times as bright as it should be, and this was 3 stops. This isn't a matter of subjective opinion, it's an objective fact that the value for that one was absolutely, completely wrong.
If we're talking variances of -/+ 0.5 EV or so sure, that won't make a massive difference. But 3 stops is 8x the brightness.
This depends on your goals. If you're not trying to replicate a material faithfully, I agree. But even then, let's say you're making a variation of a material with a different color for artistic reasons, it's still better to have an accurate source. Otherwise you're making a variation of a variation. Just like it's typically better to use photo reference when painting than looking at someone else's painting.
If you're building a library of accurate scan data, getting the values right is certainly important.
What would you set the camera settings for when doing macro shots?
Also for both Aperture Priority and Manual - Flash comp is currently set to Fill-Flash. The option to turn off flash is unavailable for Manual and Aperture Priority; so I am a little confused.
If you're using a tripod, you can get away with much slower shutter speeds, basically you don't need to worry about the shutter speed if it's a steady tripod. You should use a remote trigger in that case (wired or wireless) or simply use the 2 second shutter delay to ensure that you're not causing vibrations when pressing the shutter button.
For macro, probably F5.6-8, but again it depends on what sort of object. If it's a flat wall, F4 is probably fine, if it's something with a considerable amount of depth, you may want to use a much smaller aperture like F11. Macro is pretty vague too, it depends on distance to subject, the closer you get, the narrower the DOF is.
The problem with using smaller apertures (large fstop numbers) is that if you go too small, diffraction becomes an issue and the images will start to blur. Diffraction happens when you have very small aperture openings, the best way to explain is that it's like looking through a screen door. Generally, you want to stop down to get enough in focus but no more than that. Diffraction starts to be an issue at around 5.6 on your your camera, not enough to worry about, but it will get worse as you get to 8, 11, 16, etc.
2) Is there no other way than to tape to the camera itself?
3) What specific battery should I use with the Rx100 4? I'm looking to buy extra backups. Should I just buy Np-Bx1's?
I have to admit from what I see in Unreal 4 engine it works pretty fine and perhaps you are right there. At least on not very complex scene I tried.
Still in the engine I am working for currently tweaking textures is much more easier than lighting which is have to be set up in off line, very inconvenient manner. The lighting tweak is like trying to cover things with short blanket. once you get things right in one place, they are getting off in another. Mostly because of absence of true indirect illumination and too much of a contrast everywhere.
So in my experience you have to tweak both: lighting and materials even if they promise to be physically correct. I have never yet met a situation when they started to look perfectly ok out of a box just because they are calibrated.