Відео

You implement specs just like I do - by passive-aggressively ignoring them. But jokes aside, I love your channel and the projects you do. Keep up the good work!

Yeah I think I mentioned it on your stream but I did an rudementary font renderer/parser no curves just lines. Legit took me a week or two. It's not just you who got the shapes upside down and pretty much all the issues you encountered.

Awesome video! Have you committed to trying content full time yet, or are you still just on a break? If on a break, best of luck trying to go full time on this!

Another banger. Btw, I was wondering, what theme do you use in your terminal / vim / nvim? Looks pretty good on the eyes.

I love your videos man, You've been a great inspiration

Bro just found the most complicated way to do min(distance(uv, p)).

Sebastion lague made a video on this a little bit ago, it was a very interesting topic. Never realized how much goes into the rendering of fonts

i swear to god font files are a mess. ttf, otf, woff, and woff2, all of them are a pain in the ass. i'm working on a font manager, and trying to get all data from the font files is just painful.

off topic but just wanna say I love having your videos on as background noise for while I program, they are oddly very motivational lol

Does this example assume ASLR was disabled so that the function address would be the same on each run 🤔

WIP code up at github.com/sphaerophoria/font-rendering/blob/master/src/main.zig Some extra proof that the version of the code that returned A == 1425 is correct. If we run that for all letters we see what seem to be sane offsets with some form of semi-sane incremental movements/glyph table ranges glyph index for A is 1425 (233324..233424) glyph index for B is 12 (1636..1836) glyph index for C is 13 (1836..1996) glyph index for D is 18 (3204..3328) glyph index for E is 22 (3896..3988) glyph index for F is 31 (5556..5636) glyph index for G is 32 (5636..5840) glyph index for H is 37 (7036..7116) glyph index for I is 39 (7252..7336) glyph index for J is 48 (8924..9052) glyph index for K is 49 (9052..9140) glyph index for L is 51 (9252..9308) glyph index for M is 56 (9684..9784) glyph index for N is 57 (9784..9864) glyph index for O is 63 (10780..10940) glyph index for P is 75 (13756..13884) glyph index for Q is 77 (14028..14212) glyph index for R is 78 (14212..14392) glyph index for S is 82 (15088..15312) glyph index for T is 87 (16512..16576) glyph index for U is 91 (17032..17208) glyph index for V is 102 (19884..19952) glyph index for W is 103 (19952..20076) glyph index for X is 108 (21088..21184) glyph index for Y is 109 (21184..21264) glyph index for Z is 114 (22032..22120) glyph index for [ is 623 (122028..122100) glyph index for \ is 580 (115536..115584) glyph index for ] is 624 (122100..122172) glyph index for ^ is 1402 (230796..230872) glyph index for _ is 600 (118576..118636) glyph index for ` is 1418 (232540..232604) glyph index for a is 118 (22768..23032) glyph index for b is 129 (28548..28772) glyph index for c is 130 (28772..28920) glyph index for d is 135 (29848..30072) glyph index for e is 139 (30856..31028) glyph index for f is 150 (33464..33572) glyph index for g is 151 (33572..33980) glyph index for h is 156 (35988..36096) glyph index for i is 158 (36228..36368) glyph index for j is 169 (38116..38248) glyph index for k is 170 (38248..38340) glyph index for l is 172 (38456..38548) glyph index for m is 177 (39088..39292) glyph index for n is 178 (39292..39424) glyph index for o is 184 (40408..40536) glyph index for p is 198 (43516..43700) glyph index for q is 200 (43864..44056) glyph index for r is 201 (44056..44184) glyph index for s is 205 (44688..44848) glyph index for t is 211 (46216..46336) glyph index for u is 215 (46868..47008) glyph index for v is 226 (49440..49508) glyph index for w is 227 (49508..49604) glyph index for x is 232 (50188..50284) glyph index for y is 233 (50284..50412)

By coincidence I just found a project somebody made that implements a disassembler for an 8-bit CPU as a font using all these crazy font rules!

2:02:00 unless I'm missing something that was a difference of 100, not 0. Another victim of doing live math. Weird for "A" to be so far out but it also is in the glyph table somehow.

Fonts are one of those things that you go like: "ahit is just some text in a screen that is probably easy as fuck, we discovered how to do this like in the 90s" and then you regret everything you said

here is a "fun" fact: the PDF standard has still not been updated with support for continously variable fonts, and a bunch of those have been shipping with Windows/Office since like 2018... I found out the hard way myself, also there is a bug in Adobe Acrobat (a bug in Adobe software? shocking, I know...) where PDFs with custom fonts embedded as vector graphics are not rendered correctly (which is the only workaround for the standard not getting updated).

Also "Ubuntu" font supremacy.

As someone who watched a dense 2 hour documentary on how font rendering works. I wish you luck, it fucking sucks lol

zig's reader has functions to read values with a given endianness instead of doing bigToNative

me: please tell me how to parse your file format also me: k but don't use words to do it

First

To stutter and speaking with lightning speed is not a good combination. And telling every minute not knowing something do not help to create confidence at the audience that this video may be helpful. You lost me after 10 minutes....

2:19:44 the way you can see the individual cones reminds me how people created drawing programs in scratch or something by just drawing a bunch of circle sprites when you click lol

I'm all for abusing the 3d pipeline but this is domestic violence. 😨

You have width * height pixels*number of propagation steps and only n points depending on your stroking technique. I don't know how many but I assume less so if you need a distance field just send all the points in a buffer to the vs and evaluate all the points for every pixel. You can also have your data in buckets based on the screen position, bucket[point_x % bucket_width, point_y % bucket_height], in the vs you would classify that pixel in a bucket and you would evaluate the distance to all the points in it. Empty buckets get closes points from neighbors. If all of them are empty you expend the search.

You can rasterize the distance field as a depth map with a specific set of instancing of cones and tents. Search for the publication "Fast Computation of Generalized Voronoi Diagrams Using Graphics Hardware"

cracked

this an awesome project.

bro. how the hell do you even learn this stuff? this is awesome.

bro reading in 2x speed

1:50:20 awesome haha

Thinking about it more, and watching the VOD back, I think that we kinda accidentally invented a "world" coordinate system. There are three transformations we were talking about a lot here 1. Take plane of [-1, -1]->[1,1] and scale one dimension to match the aspect ratio of the image it represents 2. Position the object after aspect ratio correction within the composition 3. Squish everything to make it aspect correct within the composition render 2. is essentially a model -> world transformation 3. is essentially a world -> view transformation The only odd thing here compared to when we work in 3d is that the input object has these weird broken coordinates. Since all images are rendered in the [-1, 1] -> [1, 1] plane, we have to fix them when we convert them into world space. We could probably merge matrices 1 and 2, but it might be beneficial to keep them separate. That way we can easily completely reset the transform to mean "fill a 1x1 box in world space in whatever way takes up the most space"

11:13 😂

You can reuse path code for scaling, create rectangular path for scaling and then you can get scale as <original obj dimensions>/<path dimensions>

Wide Cersi is new best meme

This guy is a freaking programming terminator!

you are doing it to yourself by having object space normalized to -1..1 instead of counting pixels like a normal person

In Soviet Russia coordinate transforms love hate you.

This seems like its going to be really good for fanbox enjoyers

What happened to the ball machine?

a cool functionality would be scaling the image (super-resolution) without degrading the image quality

Is there any prerequisite to this like i know C/C++ and rust?

Hey man, I've been really enjoying your vids. I'm shocked that your viewership and comments are so low. I don't know if maybe you could collab with Tsoding or ThePrimeagen? I think you and Tsoding would have a pretty major overlap in viewership.

Just wondering, if the "flat" nature of the project an artefact of getting to a working product fast or a design choice ? (By flat, i mean that the only textures are loaded image and the window background)

~ 1:23:00 (still watching at the time of commenting) - i was expecting an even dumber/easier/faster solution: let opengl render the polygon and use the output as the mask 😅 you could also use shaders to modify the edges and such (for feathering and whatever) right at mask creation

you always have fun projects but I just don't have the time. Sacrificial comment for the algo

Hello sir Sphaero ! It's me again, from the breakout chamber ! Unfortunately the version of the chamber you accepted was using partial rendering, as I thought it would work given the comment of the render function, but the main simulation apparently requires a full render instead. So I have submitted a new version doing a full render each frame, hope this one will work ! Regards, thank you for your nice explorative videos and fun programming projects !

Clipboard is something I've looked at briefly to implement in the future in my terminal text editor. Of course, I'll only need text ,but it seems obnoxiously complex, especially when taking into consideration multiple OS's. It's a problem I haven't seriously tackled yet, but also one I have no idea how to solve, becuase there's also security to take into consideration as well.

These streams really are exceptionally inspirational, and unusually tied to real world uses in the space of recreational programming. Thank you for all this content

Yo Yo Yo never can be there when u stream but ur vids from the streams a good, guess we have the same kind of humor, im really wondering what ur cooking become when ur done

there are far too few people creating cool shit with such wisdom, mad respect