I love that my shiny idea got taken, now that my idea to make things easier to see was taken, I have another one to roll the ease of seeing back, animated Black2/White2 Spindas. PS: The frame of the animation isn't used for what dots are best and choosing a specific animation frame to increase accuracy, no, the spindas are just cycling their idle animation so they look alive, maybe with slightly different framerate speeds and randomized starting frames so it looks more like a lively spinda crowd than a wavy jello. Bonus point for having a massive volley of randomized pokeballs spammed at the screen just before the file starts, just for that satisfaction of seeing all the spindas be summoned to do their job of being an organic screen.
It was pretty wild that at first it was hard for me to look at, but then I took off my glasses, and it all made sense. Spinda is my favorite pokemon, great video 10/10 would spinda again
Ok, this was fun the first 4 minutes and 34 seconds, but it became impossible to see what is going on... My brain just goes 'spinda no spinda pls spinda stop spinda doing spinda this spinda bit spinda'.
Considering the video's production values, I thought this was a video made by a creator with like 100K subscribers, but you have only 57! That's crazy! O_O
Are genetic algorithms really the solution here? I feel like the problem is fairly simple and linear, you can probably solve it with a linear program or a gradient based approach. I think without wasting time generating parents you could probably solve all the spinda in parallel using tensors in pytorch with a custom loss function.
@FtwXXgigady Yeah I did not use genetic algorithms but actually a "find the best matching spinda" deterministic search. My implementation is not perfect but a good tradeoff between speed and reliability
I think the main points are: - I calculate the score (=how good does it match) for each dot individually and merge the best ones to one spinda - I only calculate visible pixels, pixels that will be under other spindas get ignored - I cache the calculated best spinda for each square (=imagepart repesented by one spinda) for a few frames
sorry for the total noob question, i've never run a Python script before, how do you import the video to generate the spinda pattern? i wanted to give it a try
Hi! You need to go into the to_spinda.py and change the input path on line 29. You can just put a absolute path to you file and also put absolute paths to the other settings (just make sure they exist) and you should be fine
this is wonderful; bravo
the fact that the Bad Apple demo was more readable is incredible
I love that my shiny idea got taken, now that my idea to make things easier to see was taken, I have another one to roll the ease of seeing back, animated Black2/White2 Spindas.
PS: The frame of the animation isn't used for what dots are best and choosing a specific animation frame to increase accuracy, no, the spindas are just cycling their idle animation so they look alive, maybe with slightly different framerate speeds and randomized starting frames so it looks more like a lively spinda crowd than a wavy jello. Bonus point for having a massive volley of randomized pokeballs spammed at the screen just before the file starts, just for that satisfaction of seeing all the spindas be summoned to do their job of being an organic screen.
what are you talking about? when were pokeballs thrown at the screen?
It's a suggestion for the next version. It's up to him if he does it or not.
...This video renders a Spinda with more Spindas.
You know that you can only go deeper from here, right?
you deadass ran the whole vid through spinda lmfao
It was pretty wild that at first it was hard for me to look at, but then I took off my glasses, and it all made sense. Spinda is my favorite pokemon, great video 10/10 would spinda again
This is so silly and I love it! Also the use of shiny spindas is so clever!
😁
which ones are shiny?
@@mrosskne the green ones
This video is so wonderful on so many levels. Apart from anything else it is an incredible illustration of our patter recognition hardware.
Ok, this was fun the first 4 minutes and 34 seconds, but it became impossible to see what is going on... My brain just goes 'spinda no spinda pls spinda stop spinda doing spinda this spinda bit spinda'.
this got randomly recommended to me, and i love it. great work :)
Spindas drawing Spindas. How perverse.
That’s quite smart to add Shiny Spindas into the mix in order to give it an extra color. if only Pokémon were able to have multiple shiny variants.
That's actually really cool. Nice
now you just gotta make the audio using spinda cries :3
The funny thing about this video is that the quality of the image improves if you make your vision blurry
10:27 Bad Apple on Spinda on other Spindas.
Considering the video's production values, I thought this was a video made by a creator with like 100K subscribers, but you have only 57! That's crazy! O_O
The original Video is this ua-cam.com/video/BuV_9XW8ymo/v-deo.html, so Credit goes to him - I only did some coding 😄
it was only a matter of time before someone did this
Excellent! Really nice work. Would you please update the README to include clearer instructions on running the program?
Updated! It is very simple to use, but if you have any problems let me know!
“A lot of people say the spinda was in black and white, but really it’s more like 3 ugly shades of puke”
Lily (from HIMYM): "It hurts my eyes..."
I thought it was somehow yet another pregnancy test
holy shit this is awesome
2:10
That's a spinda²
😄
Yo this is amazing
Are genetic algorithms really the solution here? I feel like the problem is fairly simple and linear, you can probably solve it with a linear program or a gradient based approach. I think without wasting time generating parents you could probably solve all the spinda in parallel using tensors in pytorch with a custom loss function.
@FtwXXgigady Yeah I did not use genetic algorithms but actually a "find the best matching spinda" deterministic search. My implementation is not perfect but a good tradeoff between speed and reliability
@@_Dearex_ oh cool, I bet the speedups are pretty big over the genetic algorithms haha
@FtwXXgigady Oh yes. If I remember correctly they took two weeks on the short doom clip, I rendered the whole thing in a few hours
Very cool! Are there any details to your implementation that would be worth sharing?
I think the main points are:
- I calculate the score (=how good does it match) for each dot individually and merge the best ones to one spinda
- I only calculate visible pixels, pixels that will be under other spindas get ignored
- I cache the calculated best spinda for each square (=imagepart repesented by one spinda) for a few frames
If you are interested in the actual code: github.com/Dearex/spindafy
Spindaception
Dear god
15:15 is what you came for.
Yes but no 😄
I hope I can to upload actual live rendered gameplay this week
If you squint your eyes you can still see the doom gameplay
@@_Dearex_ If someone can actually play on this well it's be amazing.
you have that code available somewhere?
I have a lot of cleanup to do but i'll fork the github and upload it there
github.com/Dearex/spindafy
Code out now! Not perfectly documented but you should be able to understand and use it
sorry for the total noob question, i've never run a Python script before, how do you import the video to generate the spinda pattern? i wanted to give it a try
Hi! You need to go into the to_spinda.py and change the input path on line 29. You can just put a absolute path to you file and also put absolute paths to the other settings (just make sure they exist) and you should be fine
nice
!!
here before 5.5k views
loadsa spinda
Now with shinys!
This is amazing. I hate this.