When you randomly come across a tech channel... Dig the presentation style and watch more... Only to find one of your old papers being the topic covered in the video 🤯 The pronunciation of my name leaves something to be desired, but the rest of the explanation and delivery was something I only wish I would've had the skills or time for back then! Super well done 👏Keep up the good (and entertaining!) work
This video, particularly the edge tangent flow and line integral convolution, helped me out a tremendous amount with my job, and led to a breakthrough in a problem I was working on. This is particularly surprising, considering my work has nothing to do with video filters - but it's not an exaggeration to say this saved me weeks, maybe months, of fumbling around. Thanks mr aceroller :)
Finally, a video that explains what the D.O.G. doin'. This video fascinated me as a mechanical engineer to see a different application of all the various math tools we were taught for things like controls and dynamic systems. Everytime a new term I knew came up I got hyped. "We can use this tensor" omg yes! _Sick!_ "and then use these eigenvectors" *YYEEEEEESSS!*
Randomly changing cross hatching direction in real time every frame (or every few frames) might give you an animated look, like its being redrawn or something
@@Acerola_t yeah, i wonder if there’s a way to get that kinda worbly line effect that animated objects have when theyre redrawn, in a way that isnt super expensive
@@khing5720 maybe by selecting the cross hatch direction locally with some kind of continuous but varying noise? Varying the noise parameters could vary the "wobble" effect's intensity and frequency. Maybe you could make it totally random temporally so it looks like it's being re-drawn every frame (or every several frames)
You would probably want to lock the framerate of the changing crosshatch texture to something low, like 12 fps. Maybe make it dependent on a motion vector extracted directly (good luck) or derived from depth & normal buffers.
I can only answer this question with "Yes." Thanks to how chonky the Dog is from all those settings, especially double or.... three way dogs, you can get away with forcing a lot of interesting color choices/patterns but there is definitely a lot of fine tuning required. Like many hours of it -_-
Thanks man. Had a hard time recently to stay motivated so I started playing around with webGPU and some of your filters as a side project. I'm really having a blast with it thanks to you. Hope you are doing well, bro!
This is top tier content. I'm eternally grateful for a great knowledge you've provided Acerola. I can't stop thinking of an infinite possibilities this technique provides
i think it'd be super cool to use the vector flow field to drive the rotation for the crosshatching texture, with the first and darkest layers being parallel to the VFF and the additional layers being varied parpendicular angles rather than random rotation. This would give you a more accurate crosshatching pattern that takes form into account rather than just using raw values.
if you had a quantum computer, you could try all variations of the parameters at once, and then use the quantum reverse fourier transform to extract the one that you think looks the best using a neural network trained on your preferences
At first I thought this was just an insane joke, but then I realized this could be totally possible! We just need a NN designed for dealing with quantum sines, allowing us to connect it directly to the superposition, no need for FFT!
I love those references to Tatami Galaxy and Monogatari in the editing of the video! Also the topic is presented in a very clear way, it really makes me want to code it up myself
Love the references to Monogatari series and Tatami Galaxy! And love the use of Persona music too! The explanation is very good too, I should try watching it when my brain isn't melted due to lack of sleep though
Calling it the DOG is way funnier to me than it should be. I don't really care about shaders, but calling it the DOG kept me watching. Also loved those transitions, this is a well-made video :)
"It's kinda like the stellaris tech-tree" "Image processing / stellaris" is the intersection of topics I would've expected to see on youtube when I woke up this morning
What a banger video! Not only did you manage to reteach me complicated topics I had encountered in a more intuitive way, supported by examples, but also you manages to sustain the feel of a casual conversation throughout. What a video. Subscribed.
This is actually really helpful for people trying to achieve some stylized backgrounds for animation without going through the trouble of drawing every single one of them by hand completely. I'm currently working on an animation and have 3D modelled the entire set piece in blender so I want to turn the renders I make from them and make them look more "illustration-like" in the final work.
I'd recommend referencing the paper in my description for a more true to form description of the implementation as I do neglect mild yet important details for the sake of watchability.
First video i watched from you. The explanations are technically in depth but still easy to understand. This makes me want to watch more Videos and learn more about the topic. Great work
Bruh. All your videos slap. I have no experience in this field or real understanding of how this stuff works, but your shit makes me laugh, get curious, and wanna try it for myself. I can't support you on patreon, but it'd be worth the money if I had it.
I think my favorite was the last one, though I think there is a lot more to explore with color which can significantly elevate it for certain use cases. Regardless, awesome video as always!
despite being Not-Smart about how these worked I had a lot of fun playing around with these before the recent gshade update (rip). putting a dither behind/before them (instead of after them) with the two-point interpolation turned on made for some really interesting results!
@@Acerola_t If you set the amount of colours in the Dither pretty high you can end up with something a bit like a cel-shade oil puddle effect on darker colours and, yeah... hard same. every single preset I have ever made, entirely unusable. feels so bad.
I don't regret my follow. Literrally dying under math functions, but you answered so many questions i didn't know that they could have an answer in one go. That's amazing !
Holy shit I LOVE the Tatami Galaxy reference :) AMAZING video as always :) Have I told you how amazing you are? Damn dude you're an inspiration! Thanks :)
Alright, alright, your little hidden chainsaw man spoiler just got you my subscription 😂 And I like how interesting and informative the video is --- I haven't encountered content about these topics before!
wow. i don't understand shit yet i'm so fucking intrigued. unbelievable how much goes into something i just tick a checkbox for. we really do stand on the shoulders of giants like you.
I could totally see stuff like this being used for in-game cinematics! Like cutscenes, quick scripted action events, or just stylised screenshots. Could even go as far as a stylistic theme / visual concept for the entire game. In a game, you'd also have things like depth buffers and surface normal differences available to provide a guide for the "true edges", and I wonder if the algorithm could be adapted to use that as a form of guidance/correction.
wouldn't the amount if computation be too taxing or am I wrong. I was thinking of doing the cross hatching technique but I'm not sure if it's good for the performance.
Opening ue5 to try implemet this method, i was using normal and depth diference for edge detection but this method is waaay more stylized and cool!! TY
I am unironically applying this and your cloud scattering model to optical brain imaging. These were the conceptual tutorials my trash internet brain always needed haha.
I found your channel with your grass generation video and honestly I am so glad I did. These videos are so interesting. Can the shaders made for the final fantasy game be easily applied to unity scenes and that sort of thing?
Yeah I render all the image examples in Unity in real time technically, I just blit the images to the Unity viewport and then apply the post processing effects through the camera onRenderImage hook lol The Unity shaders are also linked in the description, but they aren't intended to be used as a production ready asset.
Acerola: as you can see, the fine detail gets preserved
Me, watching on 144p on mobile: Undoubtedly
This channel produces consistently entertaining educational videos that feel like total shitposts and I couldn't ask for more
Agreed. Instead I would like to ask for less goblin sharks.
@@PropaneWP I would like to ask for more
This channel is the difference of gaussians
Moar dog, plz.
When you randomly come across a tech channel... Dig the presentation style and watch more... Only to find one of your old papers being the topic covered in the video 🤯 The pronunciation of my name leaves something to be desired, but the rest of the explanation and delivery was something I only wish I would've had the skills or time for back then! Super well done 👏Keep up the good (and entertaining!) work
omg how flattering, your work is by far my favorite post processing effect I've implemented
sorry about pronunciations! I try my best lol
This video, particularly the edge tangent flow and line integral convolution, helped me out a tremendous amount with my job, and led to a breakthrough in a problem I was working on. This is particularly surprising, considering my work has nothing to do with video filters - but it's not an exaggeration to say this saved me weeks, maybe months, of fumbling around. Thanks mr aceroller :)
omg thanks so much!! I'm glad I could help out.
Now im very curious to know what was the work. Can you say it?
Finally, a video that explains what the D.O.G. doin'.
This video fascinated me as a mechanical engineer to see a different application of all the various math tools we were taught for things like controls and dynamic systems. Everytime a new term I knew came up I got hyped. "We can use this tensor" omg yes! _Sick!_ "and then use these eigenvectors" *YYEEEEEESSS!*
Glad to know at least 1 viewer knows what a tensor is
PROVIDENCE! it's me frank
@@user-me7hx8zf9y Frank! Wow, fancy you finding me all the way on this other side of the internet. You've always been a man of good taste.
This is the difference of gaussians
The difference of gaussians this is
Of gaussians the difference this is
This Difference is Gaussian of the
Different this of is the Gaussians
This is the difference of gaussians
Randomly changing cross hatching direction in real time every frame (or every few frames) might give you an animated look, like its being redrawn or something
I do think that's the thing that'll really drive the effect home for real time, animating it in a way that doesn't look awful
@@Acerola_t yeah, i wonder if there’s a way to get that kinda worbly line effect that animated objects have when theyre redrawn, in a way that isnt super expensive
@@khing5720 maybe by selecting the cross hatch direction locally with some kind of continuous but varying noise? Varying the noise parameters could vary the "wobble" effect's intensity and frequency. Maybe you could make it totally random temporally so it looks like it's being re-drawn every frame (or every several frames)
I'd simply make the crosshatch mask an animated loop that wiggles nicely lol
You would probably want to lock the framerate of the changing crosshatch texture to something low, like 12 fps. Maybe make it dependent on a motion vector extracted directly (good luck) or derived from depth & normal buffers.
Incredible, you made me interested in signal processing after 2 teachers bored me to death with it
If only your channel existed 15 years ago.
@@owa1985this would be too high quality for a channel of that time
@@w花bcheck Captain Disillusion, he keeps making consistently high quality videos since the rise of youtube
You are absolutely crushing the dead center of the "wildly entertaining meets comprehensively informative” Venn Diagram🙌
I feel like some of these DoGs could make a cool horror game setting, the high contrast could lead to interesting atmospheres
I can only answer this question with "Yes." Thanks to how chonky the Dog is from all those settings, especially double or.... three way dogs, you can get away with forcing a lot of interesting color choices/patterns but there is definitely a lot of fine tuning required. Like many hours of it -_-
There's a detective game called "Return of the Obra Dinn" with a similar style. I think they use dots and dithering moreso than edge detection
@@1224chrisng I really don't think Obra Dinn used DoG's except for maybe some edge detection.
The intersection between art, math, and computer science here is amazing.
Thanks man. Had a hard time recently to stay motivated so I started playing around with webGPU and some of your filters as a side project. I'm really having a blast with it thanks to you. Hope you are doing well, bro!
Thanks! WebGPU is definitely in need of some implementations of these effects
This is top tier content. I'm eternally grateful for a great knowledge you've provided Acerola. I can't stop thinking of an infinite possibilities this technique provides
Keep up the good work, i learn so much from you!
Thanks so much!!
i think it'd be super cool to use the vector flow field to drive the rotation for the crosshatching texture, with the first and darkest layers being parallel to the VFF and the additional layers being varied parpendicular angles rather than random rotation. This would give you a more accurate crosshatching pattern that takes form into account rather than just using raw values.
if you had a quantum computer, you could try all variations of the parameters at once, and then use the quantum reverse fourier transform to extract the one that you think looks the best using a neural network trained on your preferences
At first I thought this was just an insane joke, but then I realized this could be totally possible! We just need a NN designed for dealing with quantum sines, allowing us to connect it directly to the superposition, no need for FFT!
As something of a cultured nerd myself, I approve of this technobabble 👍
I love those references to Tatami Galaxy and Monogatari in the editing of the video! Also the topic is presented in a very clear way, it really makes me want to code it up myself
Same. Also the whole style is a Monogatari reference and so is the name of the channel
Love the references to Monogatari series and Tatami Galaxy! And love the use of Persona music too!
The explanation is very good too, I should try watching it when my brain isn't melted due to lack of sleep though
Calling it the DOG is way funnier to me than it should be. I don't really care about shaders, but calling it the DOG kept me watching.
Also loved those transitions, this is a well-made video :)
"It's kinda like the stellaris tech-tree" "Image processing / stellaris" is the intersection of topics I would've expected to see on youtube when I woke up this morning
idk why but the arrow with the transparent background got me better than anything else today. very silly and good comedy
crazy dedication man, not sure how many other creators have invented a time machine to improve their stuff, but you're part of a select few
god i just love these videos so much. please keep it up
What a banger video! Not only did you manage to reteach me complicated topics I had encountered in a more intuitive way, supported by examples, but also you manages to sustain the feel of a casual conversation throughout. What a video. Subscribed.
You're a seriously great communicator. I wish I had this back in my undergrad visual computing classes.
This is actually really helpful for people trying to achieve some stylized backgrounds for animation without going through the trouble of drawing every single one of them by hand completely. I'm currently working on an animation and have 3D modelled the entire set piece in blender so I want to turn the renders I make from them and make them look more "illustration-like" in the final work.
I love the pencil cross hatch! Please continue making more videos!
I'm only 2:30 in, but holy cow, you have provided so much useful information that is easy to understand. Thank you!
That's so crazy that artstyle can be defined by equations and numbers. To be able to bridge art and logic is mind blowing
Dude, your videos are amazing. Please do more of this sick image processing shit I so hated in the university. Now I begin to love it.
I've spent two days watching this video on loop with the UE4 material editor open. Thank you for this information.
I'd recommend referencing the paper in my description for a more true to form description of the implementation as I do neglect mild yet important details for the sake of watchability.
Structuring the video with the tatami galaxy sequence is great
I didn't expect to find so many sonny boy reference in one of your videos. Pure gold.
Woah, that crosshatching style looked amazing!
the math looks so legitimately legit, you almost made me do math right here.
Ok, time to playlist your whole channel. Thanks for sharing this.
Finally, he’s back
yes
You are special, I enjoy how much you seem to enjoy making these :)
First video i watched from you. The explanations are technically in depth but still easy to understand. This makes me want to watch more Videos and learn more about the topic. Great work
Mad props for using the Sonny Boy OST in the background, that show slapped!
One of my new favorite channels to watch! Funny and educational
Wonderful channel! Please keep making these videos!
Bruh. All your videos slap. I have no experience in this field or real understanding of how this stuff works, but your shit makes me laugh, get curious, and wanna try it for myself. I can't support you on patreon, but it'd be worth the money if I had it.
Woah. I think that you already have the best stylization! The crosshatch really got me. Incredible work
"in signal processing" triggered my fight or flight
I think my favorite was the last one, though I think there is a lot more to explore with color which can significantly elevate it for certain use cases.
Regardless, awesome video as always!
This is some of the best content on TH-cam
I liked the charcoal pencil filter best, but it was all very cool! Thank you for another visually and mathematically interesting video!
here again because my computer vision course brought up the difference of Gaussians, this vid is unironically super helpful for me now
that intro is really good
I don't have enough words or actions left to praise you the minimum satisfactory for that high-quality content. Thank you ❤
despite being Not-Smart about how these worked I had a lot of fun playing around with these before the recent gshade update (rip). putting a dither behind/before them (instead of after them) with the two-point interpolation turned on made for some really interesting results!
omg I never tried a pre dither, that sounds interesting
and yeah hopefully the gshade shader cloning returns, very upsetting.
@@Acerola_t If you set the amount of colours in the Dither pretty high you can end up with something a bit like a cel-shade oil puddle effect on darker colours
and, yeah... hard same. every single preset I have ever made, entirely unusable. feels so bad.
That moment at 1:35 with the chicken has me dead. I couldn't believe it. I was genuinely shocked it worked so well and the glasses just add that spice
I'm a simple guy. I see Tatami Galaxy, I like.
Goddamn the videos just get better and better
This is like a video from 2007, I love it so much.
I really like the crosshatch and watercolor ones. Your videos are amazing!
I don't regret my follow.
Literrally dying under math functions, but you answered so many questions i didn't know that they could have an answer in one go. That's amazing !
Thanks!
Thanks!
Holy shit I LOVE the Tatami Galaxy reference :)
AMAZING video as always :)
Have I told you how amazing you are? Damn dude you're an inspiration!
Thanks :)
I can't explain why, but this video feels like a masterpiece to me. Great work man, keep it up!
This is awesome! I love how you actually go into the technical details.
Alright, alright, your little hidden chainsaw man spoiler just got you my subscription 😂 And I like how interesting and informative the video is --- I haven't encountered content about these topics before!
wow. i don't understand shit yet i'm so fucking intrigued. unbelievable how much goes into something i just tick a checkbox for. we really do stand on the shoulders of giants like you.
This is going in the Chanpions hall of stuff I listen to while drawing. I am enthralled
I could totally see stuff like this being used for in-game cinematics! Like cutscenes, quick scripted action events, or just stylised screenshots. Could even go as far as a stylistic theme / visual concept for the entire game. In a game, you'd also have things like depth buffers and surface normal differences available to provide a guide for the "true edges", and I wonder if the algorithm could be adapted to use that as a form of guidance/correction.
wouldn't the amount if computation be too taxing or am I wrong. I was thinking of doing the cross hatching technique but I'm not sure if it's good for the performance.
Currently my favorite channel
Phenomenal video, one of my favourites of yours yet!
Opening ue5 to try implemet this method, i was using normal and depth diference for edge detection but this method is waaay more stylized and cool!! TY
This and your video on the Kuwahara filter is really making me wish I had some decent game ideas to try the techniques out on
I think this might be the difference of Gaussians
Best video of the holiday so far
Every episode are rising my love to math. Thank you for quality content.
This popped up in my home feed, and I'm glad it happened.
Very clear and amusing explanation. I like the format as well. Keep it up!
wake up babe, new Acerola video
Love the videos, keep up the good work
Absolutely amazing video! Subscribed.
Your passion for this is really showing through. I'll probably subscribe soon, but I'll watch another video or two first.
That last cross hatching gives the impression of comic book style which I really like.
the blatant disregard for the arrow png transparency background in a video about digital signal processing applied to images conquered my heart
Thankyou so much you intelligent gift to us editors and coders. This is ammazing
why the hell this channel isnt a 100k subs one already
Awesome and hilarious explanations of image processing! Congrats!
I am unironically applying this and your cloud scattering model to optical brain imaging. These were the conceptual tutorials my trash internet brain always needed haha.
No one else on youtube get's me to cargo init faster than you, thank you.
"in order to start stylizing images" you're telling me this won't help me procrastinate on my computer vision project?
I don’t know how I got recommended this video, but I’m glad 👌🏻
i first watched this really high and the editing made me think i was dying. 10/10
Insane amount of work you did here. Thank you
On top of the content itself, I always love your title cards!
woah ok so forget the title card the crosshatching blew me away
the crosshatching look is really gorgeous imo, especially with color
Ey yo
That monogatari opening tho
Nice
I love your channel! It's so useful as a learning tool to improve my shading ❤
8:30 I would have never guessed how many steps had been made before the dithering effect.
no idea why youtube showed this video to me, but it was fascinating-- great stuff!
I didn't know this was a visual guide to Difference of Gaussians Ultimate.
Oh, a dogged fellow, aren't we? Rewinding four times!
I found your channel with your grass generation video and honestly I am so glad I did. These videos are so interesting. Can the shaders made for the final fantasy game be easily applied to unity scenes and that sort of thing?
Yeah I render all the image examples in Unity in real time technically, I just blit the images to the Unity viewport and then apply the post processing effects through the camera onRenderImage hook lol
The Unity shaders are also linked in the description, but they aren't intended to be used as a production ready asset.
This is like a real-life Rockwell Confabulator video.