Programmer @NightdiveStudio :::. 3dmath.xyz :::. C, Vulkan :::. EDG32 palette :::. deity of 7, heptagons, and dodecahedra
Joined June 2011
- Tweets 53,908
- Following 292
- Followers 21,925
- Likes 56,931
7,448 Photos and videos
Interesting, these latest upgrades to pep I'm testing seem to compress better and faster than all formats (even max-JPEG XL).
How does one advertise that their format is the best pixel art compression format on the planet without sounding egotistical? lmao
Maybe a YouTube video??
3
13
1,982
While testing I stumbled upon a way to match palette size to the right model, which skips the "test every model" code - and I found it compresses low-res (<=16 colors) even better in almost half the time!
The 256-color images take a slight penalty, but I think this is worth it 🤔
On a whim I emailed my old Calculus professor last week (he's the one who told me about Markov Chains), showing him my pep code and the ideas behind it.
He just replied and suggested I look into "Context-Mixing" and "Matt Mahoney's PAQ"
I now know what I'll be doing this weekend
5
1,075
On a whim I emailed my old Calculus professor last week (he's the one who told me about Markov Chains), showing him my pep code and the ideas behind it.
He just replied and suggested I look into "Context-Mixing" and "Matt Mahoney's PAQ"
I now know what I'll be doing this weekend
Turns out it's only effective for small-palette images. The 64 color images actually got worse, so I had to make it multi-pass; checking which 2D-tapping model compresses the best. Slightly slower, but <=16-color images compress a LOT, especially bitmaps.
Tool link below:
2
2
58
4,584
After a bit of research it seems to be bigger-O than PPM; a lot more to compute and predict, so it'll probably be a lot slower decoding.
I'm not sure I want to make it slower for a more-than-likely minimal size decrease... I could make it an option but I wanted pep to be minimal!
249
thought I'd post this mid-June because I don't know exactly when, but I started my game development journey 20 years ago!
I was 9, my Dad showed me GameMaker 6, and I was hooked. I wonder what younger me would think of where I am today...
never give up! never stop learning!
2
33
1,199
It's been so cool spending time with Thief: The Dark Project Remastered!
We cannot wait for you all to play this; like every other Nightdive project, this has so much love and care poured into it.
The Keepers call to you now… will you answer them?
No longer a secret, Thief: The Dark Project Remastered returns!
3
3
50
2,811
Turns out it's only effective for small-palette images. The 64 color images actually got worse, so I had to make it multi-pass; checking which 2D-tapping model compresses the best. Slightly slower, but <=16-color images compress a LOT, especially bitmaps.
Tool link below:
This Saturday evening I was thinking about the pixel compression format I made, and had a stark realization that I restricted the prediction encoder to 1D and not as a 2D plane!!
Quick test just now compressed further from 849 bytes down to 545 bytes 😳 It finally beats JPEG XL
3
46
7,050
pep_tool: github.com/ENDESGA/pep_tool/…
The reason why it doesn't work for high-fidelity images makes a lot of sense; they have more complexity, and tapping a 2D plane doesn't necessarily reduce complexity compared to a 1D line.
Whereas a 2-color image is the opposite, and benefits!
1
1,033
This Saturday evening I was thinking about the pixel compression format I made, and had a stark realization that I restricted the prediction encoder to 1D and not as a 2D plane!!
Quick test just now compressed further from 849 bytes down to 545 bytes 😳 It finally beats JPEG XL
I am proud to announce...
PEP: Prediction-Encoded Pixels
This format is able to compress images (up to 256 colors) smaller than GIF, PNG, QOI, BMP, in almost all cases.
A lot of passion and research (and fixation) has gone into this:
>> github.com/ENDESGA/PEP <<
7
7
193
13,375
545 BYTES for a sprite like that is so cool :')
I wanna do some more tests, then I'll clean it up and push it soon.
But I'm very happy with the results so far, I was so caught up fine-tuning the encoder that I didn't even consider the 2D plane - I guess I assumed it'd be worse...
1
13
817
since 2020 I've had a vision for a programming language that stemmed from "GameMaker but lower level", and has now evolved into "a complete redesign of C".
I've been picking away at it in my spare time for almost 6 years, and it's only recently up to the standards I expect haha
> Made Snake using HEPT - my little abstraction-engine-thing
I'd really love to expand this thing more one day soon. It's like a super lightweight C GameMaker
-> dropbox.com/s/j8vdnc34wudi96… <-
4
2
93
5,611
I went very deep down the GPU hole (which many of you may remember), and got extremely obsessed with Vulkan.
It was fun, but I ended up steering away from it because I felt there was more potential with CPU Software Rendering.
I had fun with shaders though
x.com/ENDESGA/status/1501657…
> hept :: compute hot-reloading _
it now instantly recompiles and hot-reloads the compute shader in the middle of frame cycles the moment you hit Ctrl S~
1
14
1,033
Later this year it's the 10th anniversary of NYKRA's Kickstarter and the 5th anniversary of it's initial release...
I've got a couple ideas, but as you could imagine I'm quite caught up with my studio work.
I'm thinking of maybe releasing the source code, or remaster the OST? 🤷
3
17
1,341