Fun to have such a technical post! The approach of stacking Gaussians at different radii to create a particular falloff is also used in bloom in realtime games now, though often with a more efficient technique like dual kawase blur as the basis rather than a strict Gaussian kernel per se. (Guess what I'm currently implementing in my game haha...)
I think part of the initial problem here might have been doing the bloom in a gamma space, rather than in a HDR linear space prior to tonemapping. Pixels that clip to white might represent different amounts of underlying energy. Working with PBR pipelines in linear space makes bloom look a lot more natural in modern games than the excessive bloom era of the 2000s.
Anyway, nice writeup, really cool seeing people put in the effort when it comes to compositing. Perhaps aspects of the 'classic anime look' can return outside of just nostalgia pieces... a girl can hope, anyway.
Glad you liked it! This is definitely a more technical piece than we tend to publish, but it felt like it worked for this specific topic (which is technical by nature). We learned a lot from Josh's explanation -- and it looks like he learned from yours here as well, which is amazing to see. It would be great if Spirit Jumper and the conversation around it could solve the analog-light mystery for more artists, and maybe give the look a boost in modern times.
it's a great ideal to pursue! classic anime was really exceptional in all sorts of areas from colour design to photography, and so much of that got lost in the digital transition less because it's technically impossible in digital and more because the software affordances steered towards other looks... there's no need for it to be lost. one cool approach I've seen occasionally (only in one or two places on youtube really) is to build a multiplane camera in software like Blender, and put the virtual cels through the same rendering pipeline as 3D graphics, allowing direct simulation of hdr lighting, bokeh, cel shadows and similar optical effects in a very directly physically inspired way. i think there's a *lot* of potential for knowledge to be shared between graphics programmers and traditional animators, and a lot of space to be explored at the intersection of both...
If there was an animation software that was designed with these details incorporated, it would be amazing. It doesn’t show up on YouTube much but I had to add cel shadows on each scene relative to how many layers there were (like actual cels) manually + randomization to mimic how looked under camera. All of it varies based on if it was TV animation vs film (I.e big budget vs small). The thesis of what I tried to achieve with the film was all of this is replicable digitally with an understanding of all the processes and how they look under camera - so with the right team/developers all of this could be recreated in a program!
This is really great insight actually! For the first stage of glow, there isn't a way to convert the gamma space to linear within Apple Motion (if there is, I was never able to find out how). It was only until I was doing the final film emulation in resolve where I was able to process the bloom and other light based tweaks in a linear gamma node which gave me much better results. I didn't understand why, but your explanation just clicked for me. Glad you liked the piece!
For something that happens on screen for less than a second, sometimes barely more than a second, it’s amazing how much work it took just to make (though I guess you could say that about all animation). Glad you enjoyed it!
So cool to hear about the nitty-gritty like this. I just watched Spirit Jumper yesterday because AO recommended it, and was really impressed. I'm extra impressed by this article!
Ah, great to hear we inspired you to check it out! We're very excited to print this story -- this kind of analog light is something we've wanted to cover for years, but could never learn enough about. Huge thanks to Josh for making it happen in such a cool way.
I've always had a soft spot for animation, but as an adult, I've grown more and more interested in the processes of how things are made, and this was an amazing deep dive into something that I've never really taken notice of. It's an amazing look at how the media we use affects the end result, and how little things add up to a whole experience.
The best pieces of work across all mediums are the sum of many tiny subtle parts. The best thing about digging into all of of those parts is discovering what they are that makes work feel magical
Spktra, thanks for your generous spirit! This was a thoroughly interesting write-up, and I would be so thrilled if it results in this effect making a comeback. Every time an old anime pops up, I remember how cool this lighting is and how much I miss it! Impressive work.
Fun to have such a technical post! The approach of stacking Gaussians at different radii to create a particular falloff is also used in bloom in realtime games now, though often with a more efficient technique like dual kawase blur as the basis rather than a strict Gaussian kernel per se. (Guess what I'm currently implementing in my game haha...)
I think part of the initial problem here might have been doing the bloom in a gamma space, rather than in a HDR linear space prior to tonemapping. Pixels that clip to white might represent different amounts of underlying energy. Working with PBR pipelines in linear space makes bloom look a lot more natural in modern games than the excessive bloom era of the 2000s.
Anyway, nice writeup, really cool seeing people put in the effort when it comes to compositing. Perhaps aspects of the 'classic anime look' can return outside of just nostalgia pieces... a girl can hope, anyway.
Glad you liked it! This is definitely a more technical piece than we tend to publish, but it felt like it worked for this specific topic (which is technical by nature). We learned a lot from Josh's explanation -- and it looks like he learned from yours here as well, which is amazing to see. It would be great if Spirit Jumper and the conversation around it could solve the analog-light mystery for more artists, and maybe give the look a boost in modern times.
That’s my hope! If this can inspire more stuff emulating classic cel animation, I would be stoked 👀
it's a great ideal to pursue! classic anime was really exceptional in all sorts of areas from colour design to photography, and so much of that got lost in the digital transition less because it's technically impossible in digital and more because the software affordances steered towards other looks... there's no need for it to be lost. one cool approach I've seen occasionally (only in one or two places on youtube really) is to build a multiplane camera in software like Blender, and put the virtual cels through the same rendering pipeline as 3D graphics, allowing direct simulation of hdr lighting, bokeh, cel shadows and similar optical effects in a very directly physically inspired way. i think there's a *lot* of potential for knowledge to be shared between graphics programmers and traditional animators, and a lot of space to be explored at the intersection of both...
If there was an animation software that was designed with these details incorporated, it would be amazing. It doesn’t show up on YouTube much but I had to add cel shadows on each scene relative to how many layers there were (like actual cels) manually + randomization to mimic how looked under camera. All of it varies based on if it was TV animation vs film (I.e big budget vs small). The thesis of what I tried to achieve with the film was all of this is replicable digitally with an understanding of all the processes and how they look under camera - so with the right team/developers all of this could be recreated in a program!
Perhaps it would be possible to build a package of Blender plugins for it.
This is really great insight actually! For the first stage of glow, there isn't a way to convert the gamma space to linear within Apple Motion (if there is, I was never able to find out how). It was only until I was doing the final film emulation in resolve where I was able to process the bloom and other light based tweaks in a linear gamma node which gave me much better results. I didn't understand why, but your explanation just clicked for me. Glad you liked the piece!
Thanks John & Jules for letting me write about this for everyone!
Thank YOU for the great article! Super happy we could put it out there.
How they used to animate such blinding light was a question I didn't know I had, and I'm happy to see it explained so well
For something that happens on screen for less than a second, sometimes barely more than a second, it’s amazing how much work it took just to make (though I guess you could say that about all animation). Glad you enjoyed it!
So cool to hear about the nitty-gritty like this. I just watched Spirit Jumper yesterday because AO recommended it, and was really impressed. I'm extra impressed by this article!
Ah, great to hear we inspired you to check it out! We're very excited to print this story -- this kind of analog light is something we've wanted to cover for years, but could never learn enough about. Huge thanks to Josh for making it happen in such a cool way.
Thank you so much!
really awesome, thank you!!
Oh wow. This was an eye opener for me.
I've always had a soft spot for animation, but as an adult, I've grown more and more interested in the processes of how things are made, and this was an amazing deep dive into something that I've never really taken notice of. It's an amazing look at how the media we use affects the end result, and how little things add up to a whole experience.
The best pieces of work across all mediums are the sum of many tiny subtle parts. The best thing about digging into all of of those parts is discovering what they are that makes work feel magical
Spktra, thanks for your generous spirit! This was a thoroughly interesting write-up, and I would be so thrilled if it results in this effect making a comeback. Every time an old anime pops up, I remember how cool this lighting is and how much I miss it! Impressive work.