Whether you're looking to replicate your hometown, explore urban environments, or simply build something unique and realistic, Arnis offers a comprehensive toolset to achieve your vision.
It processes large-scale data to create accurate cities, landmarks, and natural features in Minecraft, making it easy to replicate real places or design realistic environments.
(This rephrasing generated by ChatGPT)
Good technical writing is concise and tells you what you need to know as directly as possible. ChatGPT written documentation always reads like bad marketing copy.
I asked ChatGPT for a concise version and this is what it did:
"Arnis generates Minecraft worlds based on real-world locations. Select an area to recreate cities, landmarks, and landscapes in the game."
So overall it's up to the person to decide how it should look like.
So there is an existing solution in Python, seems reasonable to flag why this is different up front.
I'd be interested to understand how the buildings were constructed?
Probably easier to determine whether a verbal or (lat,long,game) location reference is a reference to an in-game location or an virtual location in a game with the same name.
From https://developers.google.com/maps/documentation/tile/3d-til... re: Google Maps GlTF 3d tiles:
> Note: To render Google's photorealistic tiles, you must use a 3D Tiles renderer that supports the display of copyright attribution. You can use [CesiumJS or Cesium for Unreal]
Open source minecraft games with Python APIs:
> sensorcraft is like minecraft but in python with pyglet for OpenGL 3D; self.add_block(), gravity, ai, circuits
WebGL is more portable than OpenGL; or can't pyglet be compiled to WASM?
panda3d and godot easily compile to WASM FWIU.
An open source physics engine in Python that might be useful for minecraft clones: Genesis-Embodied-AI/Genesis; "Genesis: A Generative and Universal Physics Engine for Robotics and Beyond" https://news.ycombinator.com/item?id=42457213#42457224
From "Approximating mathematical constants using Minecraft" https://news.ycombinator.com/item?id=42319313 :
> [ luanti, minetest, mcpi, MCPI-Revival/minecraft-pi-reborn ]
In case of Minecraft, you do have Redstone and Command Blocks at your disposal, and people have built things like an 8086 or Tetris. I wonder if there's an edge case with those WG tools, where the generated world becomes GPL-contaminated.
(Personally I'm a fan of BSD/MIT licenses, for those kinds of reasons - I'd rather not have anyone pay a lawyer to answer a simple question like, "can I actually use this program?")
It's probably not as detailed though.
https://youtube.com/watch?v=vl_MExz52jA
And the same plus some explanation from a (fake?) child:
If anyone could post a zip file of Minecraft worlds so I can fool this app into using it, I would be much obliged.
Just to give some perspective: More than 10 years ago I built an open source project that generates cities for a different game (maps4cim [1], a map generator for Cities in Motion 2). It relies on OSM data to generate roads and buildings and stuff, and NASA SRTM data for elevations. The OSM part of the application is far more complicated, since it covers so much data that is changing every day, yet you can make the exact same query against the Overpass API and get a response with the latest data in the expected format. NASA however at some point decided to shut down public access to SRTM data, which is in the public domain, and hide it behind a non-RESTful webservice where only registered users have access.
To me this feels like such a loss. I never would have started this open source project a decade ago if there was no free access to all the data the users would need. Yes, it's still public domain data, therefore someone could go through all the hassle of signing up for the service, scrape all the data and then host it free of charge, but right now there is no such service. Don't get me wrong, I don't feel entitled to this kind of free service, but I feel like these barriers to entry really matter and we all lose access to great tools that are built on the shoulders of giants. Very sad to lose NASA as one of the giants that once gave us free and easy access to highly useful data. More power to projects like OSM that live the spirit of open data!
https://www.reddit.com/r/gis/comments/qg4b6j/is_it_possible_...
Also I know Geonames has a service for elevation lookup based on SRTM & other data: http://www.geonames.org/export/web-services.html
But if you need to semple a big are in detail, that could potentially run into API limits or overload the service. But it demonstrates its possible to get the dataset & provide an API on top of it by a third party. :)
https://portal.opentopography.org/raster?opentopoID=OTSRTM.0... allows to submit a request for data from a specific area through a web form (no REST API), then you get redirected to a page where you get a download link after your request has been processed. Does not give access to the original SRTM data.
http://www.geonames.org/export/web-services.html exposes SRTM data through a REST API in their own custom format. I immediately ran into rate limiting when I tried to download a small map tile.
https://download.geoservice.dlr.de/SRTM_XSAR/ provides free access to SRTM data in the original format. This is a great resource! Unfortunately, the map tiles cover 10x10 degrees, which results in huge files. NASA used to provide 1x1 degree tiles, which was much better suited for on-demand access.
https://download.geoservice.dlr.de/TDM30_EDEM/ DLR does provide 1x1 degree tiles for similar data sets like TanDEM-X, but for those they do not provide anonymous access.
That being said, the fact that there's always something to fix just about anywhere I go tells you there's not enough people that contribute. Tbh I think the major problem is that there's no single go to app that people use all the time and then notice the issues. Though that's changing a bit with Organic Maps. Tesla self parking also had a bunch of new people mapping places, or so I heard.
Edit: I dug deeper into the OSM stuff. Apparently the building is rendered in OSM as a "multipolygon" with two different layers; but all the other stuff is rendered based on sets of nodes.
- https://streets.gl (data outdated, sadly)
- https://demo.f4map.com/ (commercial, no textures)
- OSM2World (sadly not interactively on the web, good textures, lots of rendered details, e.g. street signs)
- osmbuildings.org (no live data, no roof shapes)
None are really photo-realistic, and there's are a bunch of problems in the map data where it's hard to infer good 3D geometry, e.g. stairs.
At first I was like damn, he made a way to interact with the world map, but apparently there is a library for that! Who would have known.
Cool project nevertheless.
I wonder about the performance claims though.
> Performance Optimization: Utilize Rust’s memory safety and concurrency features to optimize the performance of the world generation process.
Is not much of an answer of performance.
What were the limitations in Python, and how did Rust help?
Since it relies on third-party libraries, what are the optimizations that the author did indeed implement as opposed to the Python version?