squaduck

[squaduck]

"fruit has sugar" warning post reminds me of my coworker who told me to make sure I don't get "addicted to fruit". yeah i'm also addicted to a nice walk on the beach

[squaduck]

Happy Stroganoff Sunday

[squaduck]

how do you make heavy ice?

you use heavy water.

[squaduck]

The op turned off reblogs but I needed to share this so bad. It’s music summer

[squaduck]

Do not prosive me

Nope. Gonna perceive you. Sorry!

[squaduck]

FUCK YOU FUCK YOU FUCK YOU FUCK YLU FU K YOU FUCK YOUNFUCK YOU FUVK YOU FUVK YOU FUCK YOU EXPLODING CAR HAMMERS DOX & STALK PHOTOMATT CHASING TRANS WOMAN TO ANOTHER WEBSITE "UNWARRANTED AND HARMFUL" WE DIDNT START THE FIIIrE

[squaduck]

[squaduck]

[squaduck]

There is such a thing as too much Unicode

I really want to explain how it works, but until I learn how to mentally project 3d gpu arrays into a physical form I can't do it justice. I also don't want to make it open source until the source code isn't awful. A real lose-lose situation. Clearly I need to rewrite it in rust once I'm sufficiently competent in rust.

Slightly sleep deprived explanation:

Basically, the core idea of how it works is this: given an image to process and an image of each character, we consider each region in the image where a character is going to be. We total the linearized rgb values for the entire region, and then calculate the total of the region after it has been masked by each character bitmap. We can subtract the total total and the masked total to get a background total. Then we divide by the area of the bitmap (or the background of the bitmap) to get linear foreground and background colors. Finally, we select the character that has the greatest euclidean distance between its fg and bg colors.

Maybe that makes sense.

I have it optimized to the point where it can render simpler (more regions of constant color) smaller videos with less available characters (~200) at ~100fps. I would give exact metrics, but it's 3:30 AM and I'm typing this in bed on my phone.

[squaduck]

There is such a thing as too much Unicode

I really want to explain how it works, but until I learn how to mentally project 3d gpu arrays into a physical form I can't do it justice. I also don't want to make it open source until the source code isn't awful. A real lose-lose situation. Clearly I need to rewrite it in rust once I'm sufficiently competent in rust.

Slightly sleep deprived explanation:

Basically, the core idea of how it works is this: given an image to process and an image of each character, we consider each region in the image where a character is going to be. We total the linearized rgb values for the entire region, and then calculate the total of the region after it has been masked by each character bitmap. We can subtract the total total and the masked total to get a background total. Then we divide by the area of the bitmap (or the background of the bitmap) to get linear foreground and background colors. Finally, we select the character that has the greatest euclidean distance between its fg and bg colors.

Maybe that makes sense.

I have it optimized to the point where it can render simpler (more regions of constant color) smaller videos with less available characters (~200) at ~100fps. I would give exact metrics, but it's 3:30 AM and I'm typing this in bed on my phone.

[squaduck]

I love having mutuals. #myfriends<3

[squaduck]

Day 147: You Can Imagine The Sound This Makes

[squaduck]

Day 149: The Logical Conclusion

[squaduck]

honestly what the hell are you supposed to do. is my question