Making Bitmaps Using a Visual Programing Language | Hacker News Reader

Making Bitmaps Using a Visual Programing Language

github.com

38 pointsbsella13 comments

In the same category, substance: https://www.allegorithmic.com/products/substance-designer

nice!

The link gives me a 502 on iOS.

Very interesting project! I'm also building something like this. I'm also using constructive geometry to build a scene, but I'm using tree hierarchy instead of graph of nodes, and primitive form of ray-tracing to produce output. Few days ago I managed to get recursion to work. Graphics programming is insanely fun!

oh sounds good, can I take a look on the code?

The code is not yet presentable, but it's open: https://github.com/jmiskovic/arbo

This is simultaneously quite cool - a tweakable image generation pipeline, like shaders; and a good demonstration of why visual programming languages remain extremely niche (Scratch, Labview): the instant descent into sphagetti.

I disagree. A textual programming language has the exact same structure. But in text. Logically a graph of nodes represented in text is harder to read then a graph of nodes represented in pictures. Most people think text is easier because of habit and because creating visual languages is harder therefore a good one hasn't been made yet.

Actually, the truth lies somewhere in the middle. Text excels at abstractice capability, relationships are less explicit via name and procedure bindings, they can be even be compressed via loops and functions. On the other hand, relationships are very much explicit and in your face in a patch/wire visual language.

This is just a trade off, visual languages are better at being explicit but worse at being abstract. When you think about it, that makes sense: a picture requires the commitment of detail that an abstract textual description does not. Incidentally, simply designing a more abstract visual notation is no easy way out of this, because the notation will inevitably resemble text as it becomes more abstract.

Also, most VPLs have some amount of textual language I them. A patch/wire VPL will go text on patch names and port labeling. Scratch, in contrast, is basically graphically edited text with some visual elements. This makes judging the entire visual paradigm even more tricky, because it isn’t homogenous at all.

What is text but a visual abstraction? The same abstractions can be made with different geometry that isn't restricted to lines and left to right reading order.

The best for me is artmatic: http://www.artmatic.com There are presets of trees that makes it easy to generate fine visuals and the primitive themselves are finetuned to generate impressive results. This is catch with that type of generators, giving the right sets of primitives, if you are too detailed it can take lomg to implemente something. Eric Wenger, the author of artmatic, is both an artist and a coder, there are not many creative peoples like him.

Cool project! Looks very similar to generating images with Blenders compositor. What are the differences between this and that?

This is a tried and true method for processing bitmaps. It is always useful to see intermediate composites as you work. Some other software implementing node-based compositing and generation:

- Nuke

- Fusion

- Substance Designer

- Houdini

- Blender

- Natron

- Autodesk Smoke

I am a fan of Houdini which gives you the ability to drop into a C-like DSL called Vex if you ever feel like tossing some traditional code instructions into the mix.

Reply on news.ycombinator.com