How To Camera
category: gfx [glöplog]
I'd like to discuss how good camerawork can be done in a demo context. In the last years, I've read a few books on cinematography and have had pixtur lecture me constantly to generally take it slower. We've done many experiments as everyweek projects and discussed them and now it feels like I went from "60 degree fov, point cam right at the thing" coder-cam to a point where I'm actually able to realize what choices I have in a particular scene. Still, I have no idea what I'm doing [tm]. Since direction and composition seem to be hot topics at the moment, maybe we can collect some insights and techniques in this thread.
Slightly off-topic, I have the impression that there are two levels of camera handling:
1. Making it look not shitty. Shots are properly framed, things look somehow slick and professional (sometimes you can't even tell why). Visual content is interesting, not everything is revealed at once. Perspective is non-confusing (or the intended amount of confusing). Things feel like they have the proper size. Camera does not feel like a weightless point in space. Aesthetics.
2. Actual cinematography. The composition uses the "language" that movies have developed over last 100 years to prime ideas in our minds, often subconsciously. If enough of those cues come together, we are inclined to follow the story more easily, get more immersed, feel more intense emotions. The most primitive example: in old western movies, there was always one screen direction (left or right) "into town" and the other "out of town". Never explicitly stated, seldom noticed, but reducing audience confusion immensely. Other examples sound more far-fetched at first, like a character in the center of the frame being lit by very strong light in an underexposed room can convey a sense of isolation, self-absorbedness, idiosyncrasy, maybe even madness (I'm thinking of a scene near the end of apocalypse now). Most of us don't consciously notice that stuff when watching the movie (or demo), but if things like this get hammered in all the time, some of them will stick and will push us into the right direction. This is where the real magic happens.
I'd be happy to reach level 1, and to be honest many scene productions don't even try. Some are definitely at level 2, but that's probably only a handful. I'd like to see more of both. Since I have no Idea how level 2 can be learned, I'd suggest sticking to the basics. I'd like to know: What lessons did you learn? What kind of controls do you use to direct the camera? How do you interpolate between camera keyframes? How do you do camera shake? How do you try to give the camera "weight" (really, this is tricky and I don't know)? How do you choose shots?
My personal "most bang for the buck" insight was: use longer lenses (that is, turn down your fov to 15 or 10 or 5 degrees). Mapping fov control to ctrl-mousewheel was the game-changer that has made us experiment a lot more. Instant level-up!
Second: linear interpolation looks okay, everything spline-y is questionable. Mostly use linear movements (maybe with variable speed, but keep cam movement on a line) and cut when you need to go to a new shot. No hermite-flying-around.
Third: look how camera is done in actual movies and try to simulate that. Put your virtual cam on a crane or a dolly and control its parameters.
Fourth: Use lots of DoF if things are small, no DoF if things are large.
Fifth: Improbably fast camera movements are forbidden, especially if things are large.
Sixth: Don't show things indiscriminately from all sides. Many objects have what Hollywood calls an "axis of action": The object's direction of movement, look direction, or similar, forms an imaginary line. Don't cross that axis with the camera while cutting, or the object will seem to point the wrong way.
What are your insights? How do you do it?
Slightly off-topic, I have the impression that there are two levels of camera handling:
1. Making it look not shitty. Shots are properly framed, things look somehow slick and professional (sometimes you can't even tell why). Visual content is interesting, not everything is revealed at once. Perspective is non-confusing (or the intended amount of confusing). Things feel like they have the proper size. Camera does not feel like a weightless point in space. Aesthetics.
2. Actual cinematography. The composition uses the "language" that movies have developed over last 100 years to prime ideas in our minds, often subconsciously. If enough of those cues come together, we are inclined to follow the story more easily, get more immersed, feel more intense emotions. The most primitive example: in old western movies, there was always one screen direction (left or right) "into town" and the other "out of town". Never explicitly stated, seldom noticed, but reducing audience confusion immensely. Other examples sound more far-fetched at first, like a character in the center of the frame being lit by very strong light in an underexposed room can convey a sense of isolation, self-absorbedness, idiosyncrasy, maybe even madness (I'm thinking of a scene near the end of apocalypse now). Most of us don't consciously notice that stuff when watching the movie (or demo), but if things like this get hammered in all the time, some of them will stick and will push us into the right direction. This is where the real magic happens.
I'd be happy to reach level 1, and to be honest many scene productions don't even try. Some are definitely at level 2, but that's probably only a handful. I'd like to see more of both. Since I have no Idea how level 2 can be learned, I'd suggest sticking to the basics. I'd like to know: What lessons did you learn? What kind of controls do you use to direct the camera? How do you interpolate between camera keyframes? How do you do camera shake? How do you try to give the camera "weight" (really, this is tricky and I don't know)? How do you choose shots?
My personal "most bang for the buck" insight was: use longer lenses (that is, turn down your fov to 15 or 10 or 5 degrees). Mapping fov control to ctrl-mousewheel was the game-changer that has made us experiment a lot more. Instant level-up!
Second: linear interpolation looks okay, everything spline-y is questionable. Mostly use linear movements (maybe with variable speed, but keep cam movement on a line) and cut when you need to go to a new shot. No hermite-flying-around.
Third: look how camera is done in actual movies and try to simulate that. Put your virtual cam on a crane or a dolly and control its parameters.
Fourth: Use lots of DoF if things are small, no DoF if things are large.
Fifth: Improbably fast camera movements are forbidden, especially if things are large.
Sixth: Don't show things indiscriminately from all sides. Many objects have what Hollywood calls an "axis of action": The object's direction of movement, look direction, or similar, forms an imaginary line. Don't cross that axis with the camera while cutting, or the object will seem to point the wrong way.
What are your insights? How do you do it?
Good tips. Also, seventh: don't point your camera directly at the thing you want to show. It's visually more interesting if it's not in the absolute dead center of the screen.
When I was setting up Signal Lost, I had this imagine in mind from the first Alone in the Dark game when Edward Carnby is coming into the Derceto manor and there's a shot where he's seen from inside a window with two hands on the window frame. So I started running with the idea that the camera in most cases is an additional (malevolent? creepy?) creature that is stalking the protagonist, and thus I usually placed it in spots where it would be either outside the field of view of the protagonist, or it'd be behind something, or in a shadow. I think in the end it ended up adding a lot to the ominous feel of the end product, suggesting that the protagonist is not alone.
This is a very good topic and a one I've been hoping someone would make for a while now. I don't have the time to go over all my thoughts right now, but there are a couple points I'd like to share. Keep in mind I'm by no means very experienced with this.
This is a supremely important point. Cinematography (i.e. the art of imaging movement) and film by proxy is largely a language of the subconscious. The most impactful moments in any film-like medium tend to be or be supported by material we don't immediately notice. It's important not to put it all out there, but to imply as much as possible.
Very, VERY important. Possibly next to the scene itself this is by far the most important aspect of a shot and its composition. The focal length (and I'm going to favor this terminology instead of "fov") implies a ridiculous amount about your scene. Typically the default should be a normal lens. What most people don't know though, and I too learned this only some months ago, is that it's not only the diameter of the film/sensor that dictates what is "normal", but also the viewing distance. In photography the relation is usually 1:1 because photos tend to be viewed from relatively close. But for films, because they tend to be viewed from further away (also one reason why watching films from a computer monitor is a bad idea*), the normal lens is considered to be twice the diameter. In typical 35mm photography a normal lens is considered to be around 50mm, but for cinema the equivalent would be somewhere in the 80-100mm range (of course cinema lenses are actually shorter since the 35mm is typically in pulldown format and thus has a smaller area). I think this should apply to demos as well, especially since compos are typically watched from similar distances compared to cinemas. A "normal fov" for demos should probably be somewhere in the 20-30 degrees range.
This of course doesn't mean it's the only choice you should use, but it certainly works well for typical shot-reverse-shot situations and typical compositions. Like cupe said I'd also encourage longer lenses if your space affords it. If it doesn't, consider making it physically larger. Traditionally in demomaking I feel most people just settle with the scene they've got and then pick a fairly wide lens to accommodate it. This has the adverse effect of shrinking the size of the scene, even though typically wide angle lenses are considered to make everything seem larger. Of course wide angles are fine, but they're also hard, and there's a reason you don't see a lot of wide angle cinematography. One of the few cinematographers that employ it extensively is three time Academy Award winner Emmanuel Lubezki.
Couldn't agree more. This especially is something I've been paying attention the more I've fiddled around with virtual cameras.
This is accurate, but I'd go beyond that: use "DoF" whenever possible and applicable. With real cameras, especially ones with a properly sized film/sensor, it's pretty rare to have absolutely everything in focus. The depth-of-field of an image tells a lot about the size relations in the scene. Actually technically it mainly says about the aperture size of your imaging system, but our brains tend to associate it with more concrete concepts. A rule of thumb also is that if your lens is wide, go easy on the "bokeh". If your lens is normal or longer, you probably shouldn't have everything in focus, unless you're focused to infinity or otherwise very far away. In general depth of field is very dependent on your focal length. Of course you should also remember that plenty of photography and cinematography is also done with very small aperture, especially in carefully composited more still shots. The whole term "DoF" in the graphics world is confusion, since in photography when you talk about a shot having "depth of field" it means that you have everything in sharp focus. In computer graphics it's the opposite.
I pretty much agree with everything else too. The shake and especially weight are very important too, the latter of which I've only very recently started to pay attention to and don't know either how to solve issues related to it. So far I think the range and speed of the movements have served as good delimiters but even those only go so far. For shake we've so far only used fairly blunt perlin noise. It looks alright, but it's obviously not very realistic unless you manage to mask it away a bit.
One thing I'd like to bring on the table and will likely make another thread about at some point is your refresh rate. To sum it up. 60 HZ IS OFTEN TIMES VERY BAD AND VERY OFTEN MAKES YOUR PRODUCTIONS LOOK BAD. There are uses for it of course especially in fast paced productions (Emulaatio looked a mess in anything less), but you should in most situations stick to somewhere in the 24-30 Hz range. This has not so much to do with film and video typically being displayed at those rates, but more with the physical properties of our eyes and our visual sense (not going into detail here). Our brains are incredible at interpolation, and you should leave it up to them to fill in the missing pieces. Displaying something causes the brain to notice a lot more that is wrong with all timings and movement that you do. One of the most glaring examples of this I noticed in Luxe, which to me looks significantly worse in 60 Hz compared to say 30. The bottom line is that the refresh rate is also a design choice that you should be aware of. Don't automatically choose 60 Hz, just because it's what interactive graphics uses (that's a whole another thing).
Probably forgot a lot of what I had originally in mind, but that's enough for now.
Quote:
Other examples sound more far-fetched at first, like a character in the center of the frame being lit by very strong light in an underexposed room can convey a sense of isolation, self-absorbedness, idiosyncrasy, maybe even madness (I'm thinking of a scene near the end of apocalypse now). Most of us don't consciously notice that stuff when watching the movie (or demo), but if things like this get hammered in all the time, some of them will stick and will push us into the right direction. This is where the real magic happens.
This is a supremely important point. Cinematography (i.e. the art of imaging movement) and film by proxy is largely a language of the subconscious. The most impactful moments in any film-like medium tend to be or be supported by material we don't immediately notice. It's important not to put it all out there, but to imply as much as possible.
Quote:
My personal "most bang for the buck" insight was: use longer lenses (that is, turn down your fov to 15 or 10 or 5 degrees).
Very, VERY important. Possibly next to the scene itself this is by far the most important aspect of a shot and its composition. The focal length (and I'm going to favor this terminology instead of "fov") implies a ridiculous amount about your scene. Typically the default should be a normal lens. What most people don't know though, and I too learned this only some months ago, is that it's not only the diameter of the film/sensor that dictates what is "normal", but also the viewing distance. In photography the relation is usually 1:1 because photos tend to be viewed from relatively close. But for films, because they tend to be viewed from further away (also one reason why watching films from a computer monitor is a bad idea*), the normal lens is considered to be twice the diameter. In typical 35mm photography a normal lens is considered to be around 50mm, but for cinema the equivalent would be somewhere in the 80-100mm range (of course cinema lenses are actually shorter since the 35mm is typically in pulldown format and thus has a smaller area). I think this should apply to demos as well, especially since compos are typically watched from similar distances compared to cinemas. A "normal fov" for demos should probably be somewhere in the 20-30 degrees range.
This of course doesn't mean it's the only choice you should use, but it certainly works well for typical shot-reverse-shot situations and typical compositions. Like cupe said I'd also encourage longer lenses if your space affords it. If it doesn't, consider making it physically larger. Traditionally in demomaking I feel most people just settle with the scene they've got and then pick a fairly wide lens to accommodate it. This has the adverse effect of shrinking the size of the scene, even though typically wide angle lenses are considered to make everything seem larger. Of course wide angles are fine, but they're also hard, and there's a reason you don't see a lot of wide angle cinematography. One of the few cinematographers that employ it extensively is three time Academy Award winner Emmanuel Lubezki.
Quote:
Second: linear interpolation looks okay, everything spline-y is questionable. Mostly use linear movements (maybe with variable speed, but keep cam movement on a line) and cut when you need to go to a new shot. No hermite-flying-around.
Third: look how camera is done in actual movies and try to simulate that. Put your virtual cam on a crane or a dolly and control its parameters.
Couldn't agree more. This especially is something I've been paying attention the more I've fiddled around with virtual cameras.
Quote:
Fourth: Use lots of DoF if things are small, no DoF if things are large.
This is accurate, but I'd go beyond that: use "DoF" whenever possible and applicable. With real cameras, especially ones with a properly sized film/sensor, it's pretty rare to have absolutely everything in focus. The depth-of-field of an image tells a lot about the size relations in the scene. Actually technically it mainly says about the aperture size of your imaging system, but our brains tend to associate it with more concrete concepts. A rule of thumb also is that if your lens is wide, go easy on the "bokeh". If your lens is normal or longer, you probably shouldn't have everything in focus, unless you're focused to infinity or otherwise very far away. In general depth of field is very dependent on your focal length. Of course you should also remember that plenty of photography and cinematography is also done with very small aperture, especially in carefully composited more still shots. The whole term "DoF" in the graphics world is confusion, since in photography when you talk about a shot having "depth of field" it means that you have everything in sharp focus. In computer graphics it's the opposite.
I pretty much agree with everything else too. The shake and especially weight are very important too, the latter of which I've only very recently started to pay attention to and don't know either how to solve issues related to it. So far I think the range and speed of the movements have served as good delimiters but even those only go so far. For shake we've so far only used fairly blunt perlin noise. It looks alright, but it's obviously not very realistic unless you manage to mask it away a bit.
One thing I'd like to bring on the table and will likely make another thread about at some point is your refresh rate. To sum it up. 60 HZ IS OFTEN TIMES VERY BAD AND VERY OFTEN MAKES YOUR PRODUCTIONS LOOK BAD. There are uses for it of course especially in fast paced productions (Emulaatio looked a mess in anything less), but you should in most situations stick to somewhere in the 24-30 Hz range. This has not so much to do with film and video typically being displayed at those rates, but more with the physical properties of our eyes and our visual sense (not going into detail here). Our brains are incredible at interpolation, and you should leave it up to them to fill in the missing pieces. Displaying something causes the brain to notice a lot more that is wrong with all timings and movement that you do. One of the most glaring examples of this I noticed in Luxe, which to me looks significantly worse in 60 Hz compared to say 30. The bottom line is that the refresh rate is also a design choice that you should be aware of. Don't automatically choose 60 Hz, just because it's what interactive graphics uses (that's a whole another thing).
Probably forgot a lot of what I had originally in mind, but that's enough for now.
Chaos Theory also looks a lot better in 24hz than in 60/120.
Also, should this thread be expanded to encompass also editing? If so, I'd recommend getting familiar with the soviet montage theory for starters. One thing it has concretized for me is to avoid cutting to music, or in the very least only cutting to music to emphasize that connection. Otherwise let the actual shot and it's contents dictate the length of the shot.
Tipp:
The "Every Frame a Painting" channel on YT:
https://www.youtube.com/user/everyframeapainting/videos
The "Every Frame a Painting" channel on YT:
https://www.youtube.com/user/everyframeapainting/videos
Super cool topic. You're probably right about the spline interpolation, which means that I could have improved every single camera in all of my my productions. I'm feeling stupid and too tired now and gotta digest this first.
The only "rule", I can think of right now is, that subconsciously, I'm frequently pulling the camera up or down to emphasize either diving into a topic/problem/mood etc. or lifting the spirit/releasing pressure/ending a part. Frequently this correlates to music.
Assigning the FOV to mouse-wheel is a super good idea. Instant copy.
@noby: I'm a little bit hesitant about this "copy what traditional movies do, because the audience will associate it with (good) cinematography". I still don't believe that the 24FPS looks better, because XYZ. Is there any real research on that topic? I totally get that camera-positions have to be believable and should not be animated without reason. But associating technical limitations like cranes and framerates with quality... Well, I'm hesitant.
The movie with the (totally subjective) best camera-work I have seen recently was "Gravity". It didn't look like they tried to stick with linear or crane moves. Yes, the movie was about disorientation in space, but still...
One more thing: It would be quiet awesome to move this discussion from an abstract "text only" level to something more visual -- like posting YouTube-link (with seconds) off scenes with outstanding camerawork in demos and explain why you think they're good. Maybe the designers can even explain, why they did, what they did. I'm not sure, there are many examples, but it would probably help me to learn and improve my own stuff.
One of my personal favorites is this one from Numb Res. Is this good camera work? It's one a spline? Could it be better?
Ah this topic is so wonderful: One more thing just came in mind: I'm not sure about the "don't put the object in the middle of the screen". From the 24 "Every frame a painting shots" above 9 are dead center.
The only "rule", I can think of right now is, that subconsciously, I'm frequently pulling the camera up or down to emphasize either diving into a topic/problem/mood etc. or lifting the spirit/releasing pressure/ending a part. Frequently this correlates to music.
Assigning the FOV to mouse-wheel is a super good idea. Instant copy.
@noby: I'm a little bit hesitant about this "copy what traditional movies do, because the audience will associate it with (good) cinematography". I still don't believe that the 24FPS looks better, because XYZ. Is there any real research on that topic? I totally get that camera-positions have to be believable and should not be animated without reason. But associating technical limitations like cranes and framerates with quality... Well, I'm hesitant.
The movie with the (totally subjective) best camera-work I have seen recently was "Gravity". It didn't look like they tried to stick with linear or crane moves. Yes, the movie was about disorientation in space, but still...
One more thing: It would be quiet awesome to move this discussion from an abstract "text only" level to something more visual -- like posting YouTube-link (with seconds) off scenes with outstanding camerawork in demos and explain why you think they're good. Maybe the designers can even explain, why they did, what they did. I'm not sure, there are many examples, but it would probably help me to learn and improve my own stuff.
One of my personal favorites is this one from Numb Res. Is this good camera work? It's one a spline? Could it be better?
Ah this topic is so wonderful: One more thing just came in mind: I'm not sure about the "don't put the object in the middle of the screen". From the 24 "Every frame a painting shots" above 9 are dead center.
Quote:
@noby: I'm a little bit hesitant about this "copy what traditional movies do, because the audience will associate it with (good) cinematography". I still don't believe that the 24FPS looks better, because XYZ. Is there any real research on that topic?
Oh I didn't mean to say or imply that. I was kinda worried actually that what I said might come across like that. I mostly just agree with cupe that the language as formulated by film is a good starting point and can guide cinematography in a lot of demos, but that doesn't mean you need to be restricted to that. But I do think if your scenes mainly consist of actual realistic geometry and you're trying to mimic an actual physical camera, it's probably good to stick to the methods used by cinematographers. There's nothing wrong with trying to create something new in this field overall, but I don't feel most people in the scene are very capable of pulling that off. As far as the 24-30 Hz thing goes, yes I've seen actual research that strongly suggests there are benefits to frame rates in that range, but I don't think I want to bring all of that into this discussion (separate thread at some point).
Quote:
I totally get that camera-positions have to be believable and should not be animated without reason. But associating technical limitations like cranes and framerates with quality... Well, I'm hesitant.
Cupe seemed to be the most vocal so far so speak for this. I'm more hesitant like you, but haven't given much thought about it yet. If you're replicating a physical camera, then definitely go for it. Otherwise I'm not that sure about it.
Quote:
The movie with the (totally subjective) best camera-work I have seen recently was "Gravity". It didn't look like they tried to stick with linear or crane moves. Yes, the movie was about disorientation in space, but still...
Funnily enough that's been the only film cinematographed by Lubezki that I didn't really care for :). Then again my qualms are related to other matters than the camerawork for the most part.
Quote:
One more thing: It would be quiet awesome to move this discussion from an abstract "text only" level to something more visual -- like posting YouTube-link (with seconds) off scenes with outstanding camerawork in demos and explain why you think they're good. Maybe the designers can even explain, why they did, what they did. I'm not sure, there are many examples, but it would probably help me to learn and improve my own stuff.
Yes! Gotta find some examples later...
To reiterate about the frame rate: It's a design decision, similar to say choosing an aspect ratio. There isn't one option that fits all, and you need to be conscious about it and not always just set everything to 16:9 60 Hz, though both of those are fine in a lot of cases. All this has an impact on the final product.
Awesome thread guys, and some great stuff mentioned so far! I'd second the "every frame a painting" series, really great stuff. I'm also in agreement that "don't center your object" isn't a one-size-fits-all idea by any means. I will add though that if your subject is _slightly off center_ and nowhere near the actual center or one of the thirds, that tends to look really bad I _think_ all the time. I'd love to see more examples where that's not the case, though.
Another thing I don't think has been mentioned yet is the use of finding lines/curves in the scene that guide your eye towards the intended subject. Need to find some good examples of that as well.
Another thing I don't think has been mentioned yet is the use of finding lines/curves in the scene that guide your eye towards the intended subject. Need to find some good examples of that as well.
About "fifth" rule (originating from pixtur's seminar I believe): I totally agree with the principles behind, but are you guys not overusing it already?
Every single demo/intro now is featuring those dead-slow camera movements. From time to time, it's ok and serves its purpose, but all the time? Demo becomes a slideshow.
Every single demo/intro now is featuring those dead-slow camera movements. From time to time, it's ok and serves its purpose, but all the time? Demo becomes a slideshow.
The "don't put stuff in the dead center" rule does not universally apply, putting things in the center can produce some very strong images. Look at more or less every movie by Stanley Kubrick: Kubrick // One-Point Perspective (vimeo)
Also, having an overlay like this in the tool helps:
Everything inside the white box is called the "title safe" area, meaning that things critical to the understanding of the plot should happen only within, especially writing. The larger black box contains the "action safe" area, meaning that any kind of action should stay within. The intersections of the faint black lines are good starting points for positioning off-center things.
I'd recommend checking the image with this kind of overlay once in a while:
because that is what it will look like :D
Everything inside the white box is called the "title safe" area, meaning that things critical to the understanding of the plot should happen only within, especially writing. The larger black box contains the "action safe" area, meaning that any kind of action should stay within. The intersections of the faint black lines are good starting points for positioning off-center things.
I'd recommend checking the image with this kind of overlay once in a while:
because that is what it will look like :D
awesome post!^_^
The problem with splines isn´t that splines are bad - actually, a lienar moving camera pointing towards a linear moving object ends in a a camera movement which isn´t linear at all. The problem with splines is feeding them with a linear parameter which results in a movement which might work for some rubber band contraption but nothing else.
A simple yet effective trick to keep things interesting: Modifying no only the main action/movement, but also changing other parameters slightly at a slower pace (e.g. zoom, side movement, lighting, angles,...).
The "slow things down" is good for amigas etc. which have a hard time keeping up with todays tech in a decent frame rate, but todays hardware is usually fast enough for 60+ fps.
And the "24fps is better than 60fps" reminds me quite a lot of the blurblurblur-hype 3dfx created for promoting their dated products...
A simple yet effective trick to keep things interesting: Modifying no only the main action/movement, but also changing other parameters slightly at a slower pace (e.g. zoom, side movement, lighting, angles,...).
The "slow things down" is good for amigas etc. which have a hard time keeping up with todays tech in a decent frame rate, but todays hardware is usually fast enough for 60+ fps.
And the "24fps is better than 60fps" reminds me quite a lot of the blurblurblur-hype 3dfx created for promoting their dated products...
great thoughts in this thread. I'm currently very interested in trying to use VR controllers for controlling a virtual demo set with virtual cameras, something VR is perfect for (imagine, "directing" the scene from inside the demo!)
I expect being able to translate most of the basic camera work techniques I've learned over the years of home movies, wild entries and other video projects, with room for experimentation. In any case, a high enough refresh rate for the virtual handheld cameras would allow all sorts of natural camera movement recording and live control.
I expect being able to translate most of the basic camera work techniques I've learned over the years of home movies, wild entries and other video projects, with room for experimentation. In any case, a high enough refresh rate for the virtual handheld cameras would allow all sorts of natural camera movement recording and live control.
I'd love to see a VR setup for playing with scenes/cameras etc. I do wonder how much "space" you need in the real world to move around in to make that effective though :)
(replace "space" with simply space :) )
Awesome stuff guys! I couldn't agree more with most of what is being said!
My tips for as you describe "level 1":
Have velocity on your cameras, though you could probably get away with a simple s-curve so it's implicit (nice for timeline based tools).
Essentially a dolly has to start moving, pickup speed and slow to a halt. Then you can just cut the right section of this movement to get the shot you want, as the middle will be pretty much linear (an idea I still need to try...).
Add various levels of shake, shake of a car engine that the camera is on, shake of the wind, shake of a persons hands. Always shake the angle, rarely is it the position of the camera that shakes.
When doing look-ats, add a human factor, it is not bad to follow an important actor in the scene, but the camera operator will always move with a delay!
My tips for as you describe "level 1":
Have velocity on your cameras, though you could probably get away with a simple s-curve so it's implicit (nice for timeline based tools).
Essentially a dolly has to start moving, pickup speed and slow to a halt. Then you can just cut the right section of this movement to get the shot you want, as the middle will be pretty much linear (an idea I still need to try...).
Add various levels of shake, shake of a car engine that the camera is on, shake of the wind, shake of a persons hands. Always shake the angle, rarely is it the position of the camera that shakes.
When doing look-ats, add a human factor, it is not bad to follow an important actor in the scene, but the camera operator will always move with a delay!
Quote:
A simple yet effective trick to keep things interesting: Modifying no only the main action/movement, but also changing other parameters slightly at a slower pace (e.g. zoom, side movement, lighting, angles,...).
I agree with slight camera movements to provide some parallax to give the shot some depth (otherwise you could rotate the cam inside a skybox), but I don't think randomly changing zoom and lighting is a good idea. In the real world, zoom lenses are expensive and heavy and have worse optical properties compared to fixed lenses, so if they don't need the variable zoom (and if there is a fixed lens available for that level of zoom), a director will choose a fixed lens. This is especially true for shorter lenses. If we assume that we are conditioned by thousands of hours of movies and series to accept and expect this kind of images, I'd really recommend stricking to it. This is generally true of all the things that film does and while it does feel like an arbitrary restriction with digital cameras, the end result will just have a more "cinematic" look if you follow the hollywood standard. Maybe this will have changed after 30 more years of digital images, but maybe not.
Arbitrary lighting changes can be confusing, since it signals that a light source is moving. If there is no "motivation" for the light (like a lamp) that could plausibly move, we add confusion to the scene which is not what we want most of the time.
Generally, your suggestions involve using time to make the shot interesting, but in line with the title of the "every frame painting" video series mentioned above I think: A shot should work without movement. Or at least the beginning and ending of the shot should.
Quote:
Generally, your suggestions involve using time to make the shot interesting, but in line with the title of the "every frame painting" video series mentioned above I think: A shot should work without movement.
Did you check episode about Akira Kurosawa in this series? They are good example how motion of the characters/background plan/weather can actually help to make the scene more intriguing, convey emotions etc... I kind of like this "cut on motion" approach as well.
A technique which is used all the time in movies but rarely seen in demos is switching between overview and close-up to give the viewer a simultaneous sense of the context and details of the same scene.
Hydrokinetics is a good example of using this technique in a demo.
Hydrokinetics is a good example of using this technique in a demo.
ive always found this difficult, the whole camera thing. i read some books on the subject too ("the grammar of the edit", good read). movies are not demos - a lot of movie editing & camerawork "rules" are focused on shooting people, closeups of people, multiple people talking and so on which are a rarity in demos. a lot of movie camerawork is heavily limited by physical restrictions which you're happily free of in a demo.
over the years ive picked up a few basic tips, like:
- use an initial shot which is zoomed out to establish the scene, then progressively come in closer to the action
- if you do a series of rotating camera shots, don't drastically change direction between them - try and make it continuous
- very slow small movements or changes in zoom can look better than totally static shots
- some of the things about shooting people can be applicable - like headroom, space in the look direction, the third wall etc
- dont overuse dof - use it heavily for extreme closeups
- break all of those rules if applicable
24 fps looks more "filmlike", when accompanied by a load of motion blur and perfect (temporal) AA, because we're familiar with it in film. we don't always have to emulate film.
i used to really worry about this but then i used to make a lot more flybys. :) nowadays i just try and focus on showing what i want to show, framing it nicely and making it look good and make sense. the camera is just a servant of the content, not the content itself ("the camera is not an effect"), so that would put me squarely in the #1 category.
over the years ive picked up a few basic tips, like:
- use an initial shot which is zoomed out to establish the scene, then progressively come in closer to the action
- if you do a series of rotating camera shots, don't drastically change direction between them - try and make it continuous
- very slow small movements or changes in zoom can look better than totally static shots
- some of the things about shooting people can be applicable - like headroom, space in the look direction, the third wall etc
- dont overuse dof - use it heavily for extreme closeups
- break all of those rules if applicable
24 fps looks more "filmlike", when accompanied by a load of motion blur and perfect (temporal) AA, because we're familiar with it in film. we don't always have to emulate film.
i used to really worry about this but then i used to make a lot more flybys. :) nowadays i just try and focus on showing what i want to show, framing it nicely and making it look good and make sense. the camera is just a servant of the content, not the content itself ("the camera is not an effect"), so that would put me squarely in the #1 category.
Quote:
- use an initial shot which is zoomed out to establish the scene, then progressively come in closer to the action
This can also be reversed to gradually reveal the whole. Or mayne it's already too much of a cliché and just corny and predictable?
This is an interesting thread, not much I can contribute but I'll be reading carefully.
