ahhh i could hear you talk about cinematic lighting all day long gleb. primarily because your credibility is undeniable. every render you have ever made (atleast everything that i have seen) seems backed up by such beautiful lighting. everything you say and teach is backed up by beautiful examples. i loved this video. for anyone who hasn't taken the cinematic lighting course, go buy it. NOW! This guy knows what he's talking about!
You know, I often doubt my understanding of these things, having a bit the imposter syndrom symptoms I guess. Or simply because I keep learning and naturally, re-evaluating what I have said and shown before, finding all sorts of inconsistencies if not simply put factual errors. I really appreciate your confirmation that at least some part of what I put together, in this format, is backed up by the visuals and feels like it's not some bs :)
@@GlebAlexandrov really interesting response. If we can do something well, does it matter if we 'really' understand it? Granted, if you're trying to teach this thing, then yes, some insight has to be there. But so much of what we call 'our thinking' is actually subconscious and intuitive. It's not something that can always be broken down into methodology, so it's not surprising that you're constantly re-evaluating your work, and your thoughts. Although, I'll admit that the concept of being really good at something is a bit subjective.
Timeline: 00:00 - Start 01:35 - What does 'cinematic outdoor lighting' mean? 03:19 - Cinematic exterior lighting in 3d - the formula 04:22 - Hidden Alley 07:02 - Forest: Aerial Shot 09:10 - Robo: Close-up 11:14 - The Leaves Close-up 11:32 - The Car in the Forest 12:21 - Reverse Key Lighting (a Skull) 13:18 - Night Highway 14:03 - Overcast Day Highway 14:35 - Night/day parking lot 15:43 - Mountain aerial view (dappled light) 16:20 - The Wall 17:00 - The Cloudscape 17:19 - The Shop 18:17 - The lighting formula - sliders 18:56 - Desert sun 19:17 - Conclusion 20:00 - Credits
so proud to see my render at 15:17 ! thanks a lot Gleb for adding my work in one of your video tuts, I feel honored as a long time subscriber to your channel :D
Just like I'm not a cinematographer yet most of the lighting videos I watch are actually made for the people who film real stuff :D I think it's good that we source the tips from the off-tangent (or neighboring) disciplines and industries, that is like consulting with a person of a different cultural background, or say, gender, to get a different perspective. Often, fruitful!
This video is a must for people learning 3D. Love the breakdown and teaching the "fundamentals" of lighting. There's no "Put a light here and set it to this random value, and im not going to explain why or what it does". Love to see more of these
It's good to know it doesn't feel that I just skip a lot of steps, so I guess the intended focus on fundamentals is clear enough. I think that these building blocks or ingredients of lighting are definitely worth talking about again and again :)
Gleb, thank you for this video! I got a bit distanced from Blender a little bit lately, mainly because I'm focusing on scripting and unreal, but your videos are always a must watch! In my mind you are part of a super elite club where only a handful of people dwell. Thank you for the days/weeks (+ decades of experience) of work you had put in this video.
Yea we've also been exploring different things including gamedev, cinematography and what not, diversifying our portfolio so to speak, so I can relate to what you say! :) Thank you for your kind words, that means a lot of me.
Your voice is so smooth and effective in conveying passion that I would stay and listen to you even if the channel was about pharmaceutical chemistry. That said, I am one of those who, after taking your course with immense interest, would gladly subscribe to another one just on landscape lighting and large-scale scenery composition. You are a invaluable guide for many. Thank you!
Thanks, Gleb, for providing in-depth information about cinematic lightning. Reverse key light is also best for rain scenes to highlight the water drops.
You're too kind! I'm constantly learning new stuff myself, in this cinematography/lighting space, by watching the aforementioned youtube channels (Wandering DP, Rob Ellis, so on and so forth).
Very interesting topic and very interesting video/discussion. I think there is actually one hard rule: the cinematic look starts with the principles of photography and especially exposure. An aspect of this is in the list as dynamic range, but you do mention other important aspects in the video which should also be on the list. For instance, if your shaders and texturing aren't correctly set up and respond differently in relation to each other than in the real world then the photographic principle breaks and making it look cinematic using the other items on the list only brings you so far. I checked the transcript and you mentioned exposure only once, while I consider it the key starting point of a photographic -> cinematic look. About the examples, I have some feedback/ideas on some of the choices you made, see it as addition to the discussion; for the robot portrait the key light actually kills the softness. Actually, the specular settings kill the softness. A lightwrap using the fresnel reflectivity might work better in this case when there's no clear directionial light. I noticed the reflection is brighter than the sky, that's inherently impossible when simulating outdoor light. Basically as you say @17:37. The part where you discuss the course while showing the night/day parking lot shows the importance of the camera response to light. For the day version I see it tries to be exposed for the diner/restaurant, but that should also expose the cars even less, but they aren't. The cars look a bit exposed for the sky, but then way underexposed for that. While the sky would be blown out anyway when the scene is exposed for the diner/restaurant, but it isn't. The 'motivation' of the light in this scene should be exposure. What to expose for and how does that affect the other elements in your scene which have a different exposure value. For the cloudscape the renderer's volumetric shader is key deterministic in whether to achieve a cinematic look or not. The forward scatter has little to do with it, that's just the term describing the direction the incoming ray of light exits the volume. With lighting 90 degrees from the camera the forward scatter moves in the direction of -90 degrees of the camera. When using backlighting for (preferred way for) cinematic lighting it's actually the backscatter which is super-important for creating both silhouette and depth perception of your cloud. Example 8 is great.
Thanks for this - so important we humans keep thinking about and building lists of the important ingredients of cinematography. I for one am not prepared for AI to get the upper hand.
Yeah creating art is still about controlling (sometimes, in quite a totalitarian manner) every aspect of the work, every ingredient and the exact dosage of everything. Art is VERY *dense*, or densely packed, by definition, everything that is in the image should be there and nothing can be taken out (at least in theory). So yea, ai or not, we'd still need to get a good grasp of the building blocks and their interrelationships.
I work in Unreal Engine & I attempting to figure out how to apply this method, there. Having some progress but I just do not seemed to grasp the theories / comprehend the concepts to get a proper handle on lighting in UE. Great video. I’ve watched it multiple times and will do more times. Very helpful and informative in a proper applied way.
Lighting can be tricky, I can relate. If you'd like, feel free to share some screenshots, and perhaps I can offer some guidance or point you in the right direction.
I will do JUST that. THANK YOU ! It would mean a lot. P.S. - Its 12 pm here (Bangladesh) I''ll be watching your video again and working / experimenting on a scene.
Congratulations Gleb! You now deserve a DP suffix for these immensely helpful insights. And don't forget, we are still waiting for your insights on animation camera moves in Blender (I think you would do a superb job with this).
I've said this for years, the real secret to good renders that a lot of beginners won't learn until later is actually nothing to do with 3d rendering at all. It's actually just good cinematography. (You can also hide a lot of crimes with good lighting and good textures)
Totally agree. But where does good cinematography come from? It actually comes from photography. Historically and practically. It's about creating beautiful, thought provoking images with lighting, composition and subject matter. People who understand photography don't need this explained to them. People need to learn to 'think' in a cinematic way, and the best starting point for that is learn photography, with a proper camera and lenses. By that I mean DSLR or SLR, if you really want to work with film. Looking at films critically is also essential as gleb touched on.
That's a high praise, I really appreciate it! Do you want to see more of such videos, attacking lighting from the 'fundamentals' perspective so to speak? :)
Thank you! Actually your comment appeared so fast that it wasn't registered by RU-vid for a first minute or so. Where did you notice the video, via a notification? :)
Yea, an interesting technique that should be used with caution as it can generate negative pixel values thus potentially breaking the whole color management pipeline.
I really like the breackdown that you made. Honestley I like video shorter, in my opinion make sense crate long video only if you have a lot things to tal and analize, but long video could be perfect solution for big analize like this video, without interruptions or thing that we don't need. Waiting the next :-) (could be awesome a video that describe how can create a really 'realistic image', because the realty is less perfect, only one video like this on YT, from sirrandalot 'Make your render unecessarily complicated'. Realized by you could be really awesome. @@GlebAlexandrov
Make a video of how to place lights so it looks good in almost all direction. For example in an animation shot, the camera orbiting a subject. It doesnt have to be realistic. But easy to the eyes at least.
That's an interesting subject indeed. There's a bit of that type of setup in this video, it's quite universal actually: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-jrCtpmdAhF0.html Updating the direction to be aligned with the reverse-key type for each new camera angle and even using it in the animation.
Oooh, I haven't seen you in awhile, Gleb, but how exciting to see you again with more lighting awesomeness! :D Dang, I already have the lighting course; how do I get that extension pack? :( I think I'd love to try using the formula in different aspects! It's fun to see how lighting can change a scene, a whole mood, even. I'm intrigued with the fire pit scene, I never see anybody trying that type. Good challenge!
I wish we could get a proper plugin of Corona for blender, I just never found the cycles materials to look that realistic and feels like a limitation for most things although speed wise it’s much better.
Yes, Corona is good, but the differences between render engines tend to be quite marginal, unless your criteria is very specific (fur or caustics, just to give an example). I sometimes read people complaining about cycles, saying it's not as good as x or y, or sometimes saying they think it's flat out cr*p, which I personally find ridiculous. Before you write off cycles completely, there are two very important things you need to consider: 1. Your materials. Are you using really good, high resolution, high quality pbr materials? 2. Your lighting. One thing I realised immediately when using cycles, is that your lighting has to be *really* good. If the lighting setup is mediocre, the end result will look cr#p. Lighting is what breathes life into your materials, and every aspect has to be considered. Direction, spread, intensity, colour pallette. If you watch really experienced artists work, it's actually surprising how much time and effort they spend on their lighting set up, and they can seem quite complex, using many light sources, of different types. If you are confident that you have excelled in this area, but you're still not getting the results you want, then by all means, go ahead and use corona. It is, after all, highly reputed, and used by archviz professionals. But, for me, I am constantly amazed by what can be achieved with cycles. I personally think that it can give stunning results.
@@richardconway6425 I am basing my opinion as a professional in the industry for 20 years. You can’t even compare cycles to something like Corona shading, they aren’t marginally different, there is a huge difference. Professional studios don’t use cycles for rendering. And yah lighting is the main thing, but lighting is also dependent on how good your shading is and cycles is just mediocre at best in terms of realistic shading. Don’t get me wrong, I think the speed of cycles has many use cases but I wouldn’t use it for final shots when there are so many better options out there
@@Mandance that's interesting. I must defer to your experience here, because I am not a professional, and I have never used corona. I have watched videos of people discussing the performance of various render engines, and showing still images of various parts of a scene, to illustrate their strengths and weaknesses. It is true that you can see differences in specific places, but in terms of the 'overall' look, I was surprised at how similar they were. It wasn't a case of one looking better, more of just being slightly different. And of course, which ever engine we use, if we're not happy with the end result, we can do all sorts of things to improve the situation. Tweaking light settings, colour grading in 3rd party editors etc. I have honestly never heard anyone state so categorically that there was such a big difference between corona and cycles. Andrew Price (Blender Guru) did a render engine comparison test a while back, and whilst he did think that corona gave the best results, it wasn't by a huge margin. If I remember correctly, he noted that it performed particularly well with transparent and translucent objects in the scene, like glass statues and ornaments. But elsewhere, I wouldn't exactly say it was overwhelmingly better, and nor did he. In fact, I remember thinking at the time how good the cycles render looked, very close to the corona render, in overall look. Considering that cycles is a free render engine built on top of a really good, free, 3d modeller, then I would say that's quite an achievement, although I realise that's not what we're really discussing here. But, I have to say, Corona is quite expensive, so it should be good. You mentioned that professional studios don't use cycles, but that's surely because most professional studios don't use blender. If they don't use blender, then they would have no reason to use cycles, preferring instead to use Arnold, or Octane or indeed Corona. Whatever fits into their pipeline. But, as far as I know, Andrew Price and his team use blender and cycles in their professional archviz work. The beauty of cycles, is that it is native to blender, so understands whatever you do in blender, and works with adds on too. It can, for example, quite happily render a mesh with loads of open modifiers running, which is definitely not the case when exporting to a different render engine. How could a proprietary, 3rd party render engine like corona be integrated to work with blender? It uses its own shaders, so wouldn't understand blender's implementation of subdivision surfaces, nor running modifiers, nor many addons. So, it's a nice idea, but I wonder if it's even possible. Perhaps it's possible, but might not be viable, from a cost/benefit point of view. Although, I have to acknowledge, that Renderman now has an official plug in for blender, but I imagine that they spent some time working with the Blender team to get it working properly. Also, that they would have good reason to do this, whereas the corona team might not have anything like the same motivation. I wonder, do you have any views on renderman? And do you specialise in any particular type of content creation?
@@Mandance that's interesting. I must defer to your experience here, because I am not a professional, and I have never used corona. I have watched videos of people discussing the performance of various render engines, and showing still images of various parts of a scene, to illustrate their strengths and weaknesses. It is true that you can see differences in specific places, but in terms of the 'overall' look, I was surprised at how similar they were. It wasn't a case of one looking better, more of just being slightly different. And of course, which ever engine we use, if we're not happy with the end result, we can do all sorts of things to improve the situation. Tweaking light settings, colour grading in 3rd party editors etc. I have honestly never heard anyone state so categorically that there was such a big difference between corona and cycles. Andrew Price (Blender Guru) did a render engine comparison test a while back, and whilst he did think that corona gave the best results, it wasn't by a huge margin. If I remember correctly, he noted that it performed particularly well with transparent and translucent objects in the scene, like glass statues and ornaments. But elsewhere, I wouldn't exactly say it was overwhelmingly better, and nor did he. In fact, I remember thinking at the time how good the cycles render looked, very close to the corona render, in overall look. Considering that cycles is a free render engine built on top of a really good, free, 3d modeller, then I would say that's quite an achievement, although I realise that's not what we're really discussing here. But, I have to say, Corona is quite expensive, so it should be good. You mentioned that professional studios don't use cycles, but that's surely because most professional studios don't use blender. If they don't use blender, then they would have no reason to use cycles, preferring instead to use Arnold, or Octane or indeed Corona. Whatever fits into their pipeline. But, as far as I know, Andrew Price and his team use blender and cycles in their professional archviz work. The beauty of cycles, is that it is native to blender, so understands whatever you do in blender, and works with adds on too. It can, for example, quite happily render a mesh with loads of open modifiers running, which is definitely not the case when exporting to a different render engine. How could a proprietary, 3rd party render engine like corona be integrated to work with blender? It uses its own shaders, so wouldn't understand blender's implementation of subdivision surfaces, nor running modifiers, nor many addons. So, it's a nice idea, but I wonder if it's even possible. Perhaps it's possible, but might not be viable, from a cost/benefit point of view. Although, I have to acknowledge, that Renderman now has an official plug in for blender, but I imagine that they spent some time working with the Blender team to get it working properly. Also, that they would have good reason to do this, whereas the corona team might not have anything like the same motivation. I wonder, do you have any views on renderman? And do you specialise in any particular type of content creation?
@@Mandance that's interesting. I must defer to your experience here, because I am not a professional, and I have never used corona. I have watched videos of people discussing the performance of various render engines, and showing still images of various parts of a scene, to illustrate their strengths and weaknesses. It is true that you can see differences in specific places, but in terms of the 'overall' look, I was surprised at how similar they were. It wasn't a case of one looking better, more of just being slightly different. And of course, which ever engine we use, if we're not happy with the end result, we can do all sorts of things to improve the situation. Tweaking light settings, colour grading in 3rd party editors etc. I have honestly never heard anyone state so categorically that there was such a big difference between corona and cycles. Andrew Price (Blender Guru) did a render engine comparison test a while back, and whilst he did think that corona gave the best results, it wasn't by a huge margin. If I remember correctly, he noted that it performed particularly well with transparent and translucent objects in the scene, like glass statues and ornaments. But elsewhere, I wouldn't exactly say it was overwhelmingly better, and nor did he. In fact, I remember thinking at the time how good the cycles render looked, very close to the corona render, in overall look. Considering that cycles is a free render engine built on top of a really good, free, 3d modeller, then I would say that's quite an achievement, although I realise that's not what we're really discussing here. But, I have to say, Corona is quite expensive, so it should be good. You mentioned that professional studios don't use cycles, but that's surely because most professional studios don't use blender. If they don't use blender, then they would have no reason to use cycles, preferring instead to use Arnold, or Octane or indeed Corona. Whatever fits into their pipeline. But, as far as I know, Andrew Price and his team use blender and cycles in their professional archviz work. The beauty of cycles, is that it is native to blender, so understands whatever you do in blender, and works with adds on too. It can, for example, quite happily render a mesh with loads of open modifiers running, which is definitely not the case when exporting to a different render engine. How could a proprietary, 3rd party render engine like corona be integrated to work with blender? It uses its own shaders, so wouldn't understand blender's implementation of subdivision surfaces, nor running modifiers, nor many addons. So, it's a nice idea, but I wonder if it's even possible. Perhaps it's possible, but might not be viable, from a cost/benefit point of view. Although, I have to acknowledge, that Renderman now has an official plug in for blender, but I imagine that they spent some time working with the Blender team to get it working properly. Also, that they would have good reason to do this, whereas the corona team might not have anything like the same motivation. I wonder, do you have any views on renderman? And do you specialise in any particular type of content creation?
@@GlebAlexandrov Yes! What I like the most is the _amount_ of different examples you go through. Some tutorials do one or two but having many makes it easier to grasp and relate too, since one of those examples might be close to what we want to achieve. Thanks !
Blender makes this very easy. In the 'collections' workspace/editor (top right) every object has some options to the right of it. 'Show in viewport', 'show in render" being two of them. Make sure these options are visible by looking at the 'filter' button at the top of the editor. I think there are at least 4 or 5 options you can toggle to show or not. 'Show in render' has a little camera icon.
I saw a few things on yt about the making of 'The Creator', and this technique was used extensively, to make all those human like robots. It looks incredible, it's uncanny how well it worked. The film itself was praised for its special effects, but sadly, as is so often the case, not its story.