Imagine that the reason for "Avatar: The Way of Water" to have taken so many years to be released was due to the caustics finally finished rendering.

I always loved seeing the ins and outs of VFX, it's a perfect example of "if you do things right, people won't be sure you've done anything at all."

Just watched avatar the way of water and it’s definitely simulated. The light patterns changes accurately when a creature is disturbing the water surface. It’s crazy how good it looks.

omg I cant believe mark is actually doing normal stuff for the channel this is so cool. this just made corridor crew 10x better.

"Real life runs constantly in real time without ever crashing." Well, I would certainly hope so. Great video Wren.

The history of long hair in animated movies. Simulating hair has been such an interesting subject. I think it would make a pretty good video.

I think the main case where simulated caustics for an underwater shot would be more notably different than the gobo option, is when the thing catching the caustics is very close to the surface of the water. The parts of the dinosaur that are right up against the water would catch very different caustics than the parts further away, not just in brightness but in sharpness as well. If you move the dinosaur further underwater, the difference between the close parts and the far parts becomes less significant, and the result looks a lot more like the gobo.

These Wren videos are some of the highest quality videos out there. His understanding and ability to explain these concepts is so unreal. Great stuff!

it's a shame to come back to a video i used to enjoy, now knowing this company is so comfortable with art theft and is even complicit in it.

Man I missed Wren's deep-dives into very specific topics. I enjoy most of the content on this channel, but those hit particularly hard for me since he usually does either a science-y topic, or something crazy and obsessive about a sci-fi franchise... in either case, I love the effort you (Wren + the whole team involved) put into these vids.

As far as I'm concerned I saw the trailer of Avatar the way of water multiple times and each time I caught some new details but talking about Caustics, yes they were both a mixture of original shots under water and perfect cgi because one can clearly see the rays of light in the trailer but that doesn't happen in real life and the Caustics get dull as depth increases also the clarity of water is the biggest factor in play here, cleaner the water more crisp are the Caustics

Wow I can't believe wren was able to speak under water so clearly!!! Fantastic job!

When you consider if someone were to actually run that 11 year render, by the time it finishes you would probably be able to run an equivalent render in an hour with the advances in both hardware and render algorithms, probably also with some sort of AI enhancement thrown in.

I was working on that in uni decades ago. The reason caustics are hard is there's no obvious way to trace a path from the camera to the light through 3rd surfaces. Normally you trace from camera to surface and then you trace in the direction of each light to see if the patch of surface is illuminated. But if light is to come through reflection or refraction it could be coming through any path. You'd have to render the entire scene from the PoV each surface patch just to calculate 2nd level lighting. Or you could trace photons forward from the light, have them deposit on surfaces randomly like textures, and then trace the scene from the camera and hope you got enough photons. That's the slow technique that I worked on way back. You could do better with fancy sampling and compression of photons, or some differential approximation of path tracing, which I'm guessing form part of the new renderer.

Thank you for this video. I am working on learning how to make water in Unreal Engine 5 right now. I have only been in the learning to render and model space for about 2 months and I think I missed my calling. I have been a software and web applications engineer for 26 years and it doesn't bring me near the pleasure I am getting from the open world game I am building and all of the rendereing, modeling, and animation that I am having to learn to build it. Thank you for all of the great content you have and will produce. Keep up the great work and keep inspiring everyone that watches your videos. If I ever had the chance I would love to come work with you guys and learn some of the craft and tricks of the trade from you and the rest of the Corridor Crew.

Wren : finds an amazing shot Wren a few minutes later: realizes the shot was simpler than imagined Also Wren : fuck it, I wanna try the hard way and see it's worth it And that's why we love this channel

Wren: "It will take 11 years to render it!" AI: Two more papers down the line. Take it or leave it.

This is something all of the content creators should adopt in their videos " Enthusiasm " Cause when you are actually having fun doing what you're doing the viewer feels the energy and it prevents him from moving an inch. I personally enjoyed every second of this video.

Congratulations, Wren. I think this is the best video you've ever done. Well done, you did a great job putting this together

Take it this way... The movie Avater 2 is truly coming after 13 years, so maybe they shot it that time and... Waited 11 years for the render!

I love Wren's solo vids. Learnt a lot once again. Keep it up.

I love these kind of videos from Corridor it’s fun and educational and Wren does a great job explaining how or why something works as it does

I think the gobo shot can be improved just by altering the actual video used for the light. So if you want to change the light falloff, you could try and change the light gradients and shapes, and adjust the intensity and contrast of the light source.

I've always simulated water wave caustics simply by using a Voronoi noise shader in World space under a Colorizer filter in the Luminance channel. The gradient of the colorizer is then mapped to an inverse-square falloff (I use a Python node to generate the gradient knots from a formula) to give it the sharp contrast between the bright thin regions and dim falloff characteristic of these kinds of caustics. It looks very close to real, and adds zero to the render time.

Wren is honestly a great educator. Has a passionate energy, knowledgeable background, and ability to simplify complicated ideas. He really could have his own education channel and I bet it'd be huge!

6:50, you can get that, physically correct caustics in Blender as well, and Blender has had that feature for over 6 months now and it's using a much more complex and advanced algorithm than photon mapping, it's called "manifold next event estimation" and it's based on a research paper released by weta digital, so there's a good chance they've used that method to calculate real caustics in avatar 2 by using their own renderer called Manuka. Great video nonetheless.

I find it interesting that caustic filters of animation and film is like the filtering of convolution reverb on audio. You can take the room sound reflection response of something and put it on audio to simulate that a recording is taking place in a specific place like a bathroom or a tunnel.

10:36 Wow! The results are amazing and they look so real. The morphing purple sphere is deeply pleasing to the eye.

Wren emerging from the deep is a callback to Sync when the main character wakes up in the ocean.

As a professional rendering engine developer, can I just say guys, this is a REALLY good video. Well done. Also, it's cute that you think 100 hours is slow to render. r.e. Avatar, WETA can certainly render caustics of the nature you showed "for real". Their renderer Manuka has a very flexible raytracing architecture that they can tune per shot. They have also released a phenomenal number of advanced papers recently for rendering complex light-transport scenarios, so I would guess that yes, those shots proabably were fully physically-plausible caustic renders (probably also with some creative license thrown on top for good measure).

This was fascinating and I feel like contained knowledge applicable outside of just caustics. Loved the deep dive!

After diving into making a washing machine tech visualization for school with simulated water exploding out of the drum and rendering it all with VRay with both caustics and motion blur I can certainly relate. This was a lovely video to see and man, I gotta check out Photon Tracer.

That montage of all the glass renders was honestly one of the most beautiful things I've seen

4:27 OMG! ITS MARK!! So happy to see him on corridor crew!

music choice at 10:35 is amazing im so glad to see drum and bass lol cause a lot of mixes on yt have old renders

Wren is always so fun to watch. i was barely able to follow as couldnt really understand and yet he still makes it so enjoyable and easy to watch :)

The randomness on the simulated caustics is why they feel better to my eye, the gobo is very cyclical. If it was an isolated beauty shot, the simulated will make a HUGE difference, but in the Dino scene it’s probably not worth it. Really cool topic!

Wren: "surely they're not called caustics because they burn things, right?" Also Wren: proceeds to burn things with caustics (In fairness, I also didn't make the connection until he said the exact words "catch things on fire".)

These videos made by wren are some of my favorite videos on corridor

I have watched several of your videos now because I'm interested in the process and it's apparent limitations and well, all computer generated graphics. In this video I got a better sense of what you really mean when you say rendering, which helps me understand so much more. I just have one question: you seem to have a lot of fun there, which is great, but how do you guys ever have the time to get all these great videos done???

Wren: I'm gonna render this scene at 16000 samples. Also Wren: Why is it taking so long ?

4:15- that guy is SO stoked to talk about his work. That’s a rare gift, to have your work be something that excites you that much.

Amazing information on this video as always thank so much guys. 🙏😍

love that wren actually went into the ocean for the shot - true dedication

Mark casually being an intern always cracks me up. Loved the video, very interesting topic.

Brilliant video. I love how you describe the craftmanship of the CGI simulations versus the reality they emulate. It is plain to see that a rendered scene needs a CG artist to set up every detail of a single scene. Yet, many think that the reality of our universe can spring into existence without any creator. 🤔

This was very entertaining! Ryan Brucks from Epic Games made a video where he addresses the subject and his results are nothing short than phenomenal. What's missing in the "fake" version of your caustics is the fact that just as with real lenses, the caustics are volumetric. That is, they can't be approximated purely by projection, because depending on the depth, rays of light converge or diverge, just like in your lens test. Ryan is pre-computing the caustics as a volumetric texture (3D), where each slice of caustics is different and uses the distance to surface to project the correct slice. You could simulate that with multiple projectors angled slightly different, each projecting a different caustics subset and when the projections meet, you basically get the non-directional caustics that render different at different distances. Probably a small difference in the output, but you get a little closer to the real thing with virtually the same rendering time.

Love to see intern Markiplier doing the hard work

I second what a few others also said: The differences between gobo-caustics and simulated ones are probably way more noticeable once anything interacts with the cause of the caustics. In the case of water the surface. Like an underwater shot of someone breaching the surface from below. The caustics on the part of the body still underwater would change as the surface would get distorted. And also there will be bubbles the light has to go through. Same for let's say a boat passing over your underwater scene. It would not just be a shadow as another gobo layer on top of the "fake" caustics. The wake of the boat would change the caustics. Although in contrast to the first example it would probably be reasonably easy to record the caustics patterns of a model boat in a pool to simply get a better "fake" layer. Oh and my guess on why the simulated magnifying glass looked slightly different is, that you modelled a perfect lens. But real lenses, especially cheaper ones, often have slight imperfections in their surfaces. And the refractive index varies as well. There is a reason you pay several 100$ for good lenses that size.

I'd like to see how VFX artists/companies go about creating new software for tools they need which were too advance at their time of creation. You could compile a list of tools created that helped VFX make large leaps and strides in the industry.

This is such an informative video, driven by the true power of science: curiousity and willingness to experiment. I love it.

As someone who has a Bachelors of VFX, it’s nice to see how much the TECH has improved over the years! These CAUSTIC effects look incredible!

I'm researching caustics and its 3d rendering at university and I must say: this was a great video. It was very informative without being overcomplicated. Also, the message about how computer graphics is all fake is super important because it's easy for us artists to forget that cgi is entirely simulation, that, for instance, no rays or photoms are really being traced or that no particle or wave is really interacting with matter and transfering energy to it etc. I say its easy to forget because I myself did it, focusing too much on physical accuracy when in the end rendering algorithms are just mathematical models that try to replicate real phenomena (and generally from the limited pov of a camera). It doesn't matter how good they are, they will never reach reality. At most they will seem visually close enough. When we remember that, we are free. Free not just to take some shortcuts as shown in the video, but free to explore cgi as what it is, a simulation, and embrace both the limitations of innacuracy and its hidden powers as part of the artwork, as a way of subverting the laws of physics and, because of that, creating better and more powerful scenes (yet physically innacurated) and telling better and more touching stories with them.

Dude that caustic render sequence with the drum n bass was so sick

That was awesome! Loved the deep dive into this niche area!

7:53 "I know its hot, but i just wanted to... feel it" I see that Wren is getting a little bit influenced by Mark-"Not a masochist"-iplier, because that sounds a little too close to "I just wanted to see if my body can take it"

As a glasses-wearer, I've always been curious about the various types of blur. For example, without my glasses, the world is blurry, but it's not like a smooth gaussian blur, or a smeared motion blur... it's kind of like a distance blur, where edges will blur out to a certain "distance", but then there's almost a hard edge, a place where the blur cuts off. In any case, it'd be fascinating to see if renderers can simulate the blur of poor eyesight somehow.

@0:25, 90s Bill Nye vibes and i’m here for it!

I actually used this vid as inspiration after watching it the first few times and ended up using caustics in some background renders for my art and in a piece I recently did for Valentine's day I even used it to simulate an old CRT TV and it's been fun messing with caustics in Blender.

This is honestly one of the most interesting things I've seen in a while, caustics are so mesmerizing! Would totally watch if you went more in depth on other types of effects!

4:17 - Good to see that intern being put to work!

As a VFX and 3d artist i love dynamics 🧬

The first episodes of the series were of course a step into an almost unknown Middle Earth... But from that song on I loved the series and I'm watching it with joy again when season 2 is coming.

So glad that Prehistoric Planet is getting more attention, especially in the CGI department.

Damn it feels like forever since the last "Wren explains cool shit" video. I missed this.

I love this channel, caustics and GTA cube maps are things that have recently blew my mind

I love this channel! You guys help me appreciate the little things in life. We really are having a intense experience as humans thst it is so difficult for humans to recreate. So many things that go into how we see our world! 🤯✨

They've taken so much time to develop new technology for the Avatar sequels, I would not be surprised at all if they've found a more effective way of doing it "for real".

The hard-edged falloff with the gobo-based effect could probably be improved by using an area light instead of a directional light. Key thing to keep in mind is that in a real water scenario you have refracted light angling in on the object from a broad area of water surface, so the light "wraps" the object's form more. Incidentally, in games we sometimes get this effect by building the caustics into the surface material itself, rather than projecting it from a particular light. That approach also makes it easy to smooth out the falloff along the edges, since you're just putting a little bias into the directional falloff.

I was on the Avatar way of the water fimling set, costics were a real deal that took an incredible amount of development from research which came to be part of the delay. In production, costics are made computer generated, since underwater recordings were made in a tank with floating balls and special recording equipment in order to get propper lighting, but to avoid getting real life costics which would interfere on the development of the new mocap underwater suit tech, which would need propper tracing suits.

Don't forget that you can always render the gobo caustics on another renderlayer and feather it out in post. So I'd say it's not feasable at the moment :D but it's come a long way !

4:25 Is that Markiplier?! 😂

0:41 that dedication from Wren we all love to see

I really do admire Prehistoric Planet. Its visual effects, its accuracy to our latest understanding, The water and David Attenborough. Its amazing.

I actually loved this video. I've known realistic water CGI is hard (mainly 'cause of seeing how often water is the bit that looks unrealistic in lower-budget movies), but didn't realise that caustics is why - given that I don't work in CGI myself. I'm definitely gonna remember the but about Avatar... Gonna be interesting to see. I'd love to see a video about hair CGI.

4:45 I didn't even realize that was Mark at first. XD

Whoever you hired to play the role of the render engine around the 4:15 Mark on the timeline was hilarious. Mark my words, that guy is gonna go places as an actor. He hit all the Marks perfectly. I can't give any bad Marks on his performance. He was never off the Mark about how to portray the role. I feel like he really left his Mark on the video. Some people aim for funny and try too hard and really overshoot the Mark, but not him. He should start a youtube channel. Now if you'll excuse me, I have to find my pliers.

Ok, here is one for you: Using just the render engine and no compositing tricks like using After Effects and such, creating a completely accurate physical atmosphere such as the one around earth. Imagine doing a documentary in which you are approaching earth from the moon, and you need to not only simulate how the atmosphere of the earth changes in appearance as the spacecraft you are in gets closer and closer until it is orbiting at the altitude of the ISS, and during this process, you witness several sunsets and sunrises, and then once in orbit, watch a few more as well. Using just the app and the render engine of your choice, try simulating the physics of absorption and scattering through multiple layers of the atmosphere such as the troposphere, stratosphere, mesosphere, ozone, etc and the behaviour of light through each in order to create the Rayleigh scattering, albedo calculations and other accurate behaviours of light involved. While Blender does have a few (paid in most cases) add-ons to help accomplish this, I would love to see you approach it in the same manner that you examined caustics in this video. If that sounds like a fun challenge, I would love to see your approach to it. Enjoy. :)

Cool vid, very well put together, 15 min went by really quick

Wren is my favorite. He’s always so stoked and excited about his craft.

As a former photographer/video producer/physics teacher/astronomer, I am perpetually looking at light artifacts/phenomena. My favorite discovery was when I was eating breakfast and my cat started going crazy. It took while because I kept shifting my head, but the cat was chasing the caustics created by an 8 lamp chandelier refracting through my glasses. Every time I moved my head the slightest, the cat saw eight light bugs (Caustics) flying around the room. It was very funny. (probably not to the cat.)

Love, Love, Love Wren hosted videos..

Just the random montage with the breakcore is awesome

As a 2D artist, I learn so much from seeing attempts to simulate light in 3D and FX software. It really breaks down the science and the effect on surfaces so that I can recreate it in my own medium.

Please make more videos like this. A few years ago I experimented with replacing the environment texture in Blender with a 360 video to make it seem like my 1950s cg car was really moving through LA with proper reflections, but I could never find a video that was HD or filmed correctly for what I was doing. I would like to see a video about a modern approach to this.

When I used to do 3d stuff it took like 8 hours to render caustics at 640x480 The plugin was called Brazil R/S by Splutterfish for 3D Studio Max, but this was back in 2005

Man, those are very nice shots. I got goosebumps

Wren's explaining videos are awesome!! Really really enjoyed this, and I love how the explanations really make us feel that we have a fundamental understanding without being too complicated. My only nitpick is that you sort of glossed over how the renderer was developed to make caustics achievable in normal amounts of time. Like you spent the whole first half explaining why caustics are impossible, but then it turns out that octane just has a caustics setting and it all works? How did they develop that and overcome what was previously impossible? But aside from that, great video!!

Wren, I hope you see this. I love these exporations into vfx shots, I see a similar enthusiasm as Adam Savage. It cool to still have some content that has someone(corridor as well) who cares so much about sharing and exploring with an audience. Thank you

Not sure if Corridor will see this. But I can tell you that NVIDIA researchers proposed a neural network-based method for rendering caustics. Trained on a large dataset of caustic patterns to be able to generate caustic pattern in real time on new 3D scenes or light fields, without ray-tracing or photon mapping calculation. They are still in research stage. They talked about it on the SIGGRAPH 2021 conference. The paper was called "Neural Caustics: Rendering Caustics from 3D Scene and Light Field" NVIDIA sadly removed it from their archive, but maybe it is hosted on some other research paper archives. As far I recall it was very impressive back in 2021 no idea how far they got with that now.

10:26 and that my guys is how you get 40.3 Tb of hard drive filled up

jeez, i never had so much respect for the guys who have to code literal reality into their render engines

I would love more of a deep dive into early graphics engines, and how they were made/developed. Between Tron and Jurassic Park is about 11 years, but the way the technology evolved is huge!

I can't even rotate the view properly in blender and look at what you are doing man that looks real

I wish I could go to Corridor and learn how to do VFX in person. All of these videos feel so educational that I bet just being in proximity would help you learn.

Just when I needed a video to my cereal to

Corridor Crew: "And it has never crashed!" Dinosaurs: "I beg to differ."

I understood a gobo to be “Go(es) Before Optics” In any case, it goes in the imaging plane of the light fixture, whether that’s before the first lens (most commonly) or somewhere in the middle.

Hello Wren, VFX supervisor here, I think that the best difference with Gobo and actual simulation is that regarding the proximity to the surface, water interaction (splashes or surface tension) cannot be achieved with the Gobo method, so if you talk about realisim as in how it looks, they are almost the same, but realism as in it can be closer to the surface and the water interacts with the object moving underneat I would go with the sim, that would be correct time invested.

7:41 Beavis chanting "fire!" was something I did not expect lmfaoo