[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-77169":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":10,"language":11,"languages":9,"totalLinesOfCode":9,"stars":12,"forks":13,"watchers":14,"openIssues":15,"contributorsCount":9,"subscribersCount":16,"size":16,"stars1d":17,"stars7d":18,"stars30d":19,"stars90d":16,"forks30d":16,"starsTrendScore":20,"compositeScore":21,"rankGlobal":9,"rankLanguage":9,"license":9,"archived":22,"fork":22,"defaultBranch":23,"hasWiki":22,"hasPages":22,"topics":9,"createdAt":9,"pushedAt":9,"updatedAt":24,"readmeContent":25,"aiSummary":26,"trendingCount":16,"starSnapshotCount":16,"syncStatus":27,"lastSyncTime":28,"discoverSource":29},77169,"terrain-diffusion-mc","xandergos\u002Fterrain-diffusion-mc","xandergos","Procedural terrain generation with diffusion models (in Minecraft)",null,"https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion-mc","Java",545,43,16,79,0,13,28,83,39,8.93,false,"main","2026-06-12 02:03:42","# Terrain Diffusion Fabric Mod [[Modrinth]](https:\u002F\u002Fmodrinth.com\u002Fmod\u002Fterrain-diffusion)\n\n#### UPDATE: The research behind this mod has been accepted to SIGGRAPH 2026, the world's premier graphics conference! That means the research was officially peer reviewed and recognized as a significant contribution to the field. Enjoy the mod!\n\nThis is a Minecraft Fabric mod integrating [Terrain Diffusion](https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion).\n\n## Which version should I use?\n\nThree builds are available on the [Releases](https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion-mc\u002Freleases) page:\n\n**The CPU build is slow unless you are on MacOS.**\n\n| Build                     | Supports                    | Setup required                          |\n|---------------------------| --------------------------- | --------------------------------------- |\n| **Windows** (recommended) | Windows with any modern GPU | None                                    |\n| **CUDA**                  | NVIDIA GPUs                 | [CUDA + cuDNN install](CUDA_INSTALL.md) |\n| **CPU**                   | Everything else             | None                                    |\n\n> **Mac users:** the CPU build automatically uses CoreML for hardware acceleration on Apple Silicon. No extra setup is needed.\n\nUse the `-cuda` build only if you are on Linux, or have an NVIDIA GPU and prefer CUDA (may improve performance).\n\n## Requirements\n\n- Minecraft with [Fabric](https:\u002F\u002Ffabricmc.net\u002F) and the [Fabric API Mod](https:\u002F\u002Fmodrinth.com\u002Fmod\u002Ffabric-api) installed\n- Windows with a GPU OR Linux with an NVIDIA GPU is strongly recommended. CPU inference works but is very slow.\n- VRAM (GPU RAM) needed: 1.5GB\n- RAM needed: 2.5GB (May need to increase Minecraft's RAM allocation)\n\n## Usage\n\n**If using CUDA build:** First see [CUDA_INSTALL.md](CUDA_INSTALL.md).\n\n1. Download the mod jar from [Releases](https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion-mc\u002Freleases) for your Minecraft version and place it in your Minecraft `mods\u002F` folder. Make sure the Minecraft version matches.\n2. Launch Minecraft, at least once online to download the models (~2.5GB).\n3. Create a world, and select the **Terrain Diffusion** world type. Click **Customize** to set the `World Scale` (see [Per-world settings](#per-world-settings) below).\n4. The mod will search for a land spawn point near the world origin automatically. If the area around (0, 0) is entirely ocean, it may take a moment to find land. Use `\u002Ftd-explore` (see below) to scout the world further.\n\n## Exploring the World\n\nThe mod includes a built-in terrain explorer web UI. Run the `\u002Ftd-explore` command in-game; it will print a clickable link (e.g. `http:\u002F\u002Flocalhost:19801`) that opens an interactive map in your browser. Click the map on the left to open a \"detailed view\". Click the detailed view to get coordinates in the bottom left. You can also filter for certain climates.\n\nUse the explorer to scout continents, mountains, islands, and other interesting terrain before venturing out in Minecraft.\n\n## Configuration\n\nEdit `config\u002Fterrain-diffusion-mc.properties` (created automatically on first launch):\n\n```\n# Terrain Diffusion MC configuration\n\n# Inference device: \"cpu\", \"gpu\", or \"auto\" (try GPU first then fall back to CPU).\n# \"gpu\" uses DirectML on the -windows build, or CUDA on the -cuda build.\n# GPU builds default to \"gpu\" so startup fails loudly if no GPU is detected.\n# CPU build defaults to \"auto\": uses CoreML on macOS, otherwise CPU.\ninference.device=gpu\n\n# Offload inactive models from VRAM between pipeline stages.\n# Keeps peak VRAM to ~1.5-2 GB. Set to false if you have ~2.5+ GB free for slightly\n# faster generation.\ninference.offload_models=true\n\n# Validate SHA-256 for pre-existing files in .minecraft\u002Fterrain-diffusion-models.\n# Set to false if you want to provide custom models\u002Fconfig files without hash checks.\nvalidate_model=true\n\n# Port for the local terrain explorer web UI (\u002Ftd-explore).\nexplorer.port=19801\n\n# Spawn search: coarse-pixel region sizes for finding a land spawn near (0, 0).\n# Starts at initial_size x initial_size and expands by 8 each step up to max_size x max_size.\n# Each coarse pixel covers a large area (hundreds of blocks), so 16–128 is typically sufficient.\nspawn_search.initial_size=16\nspawn_search.max_size=128\n```\n\n### Per-world settings\n\nFor Terrain Diffusion worlds, click **Customize** in world creation and set:\n\n- `World Scale` (integer `1..6`)\n\nThis value is saved with the world save and affects:\n\n- how many real-world meters each block represents (`scale=1` => `30m\u002Fblock`, `scale=2` => `15m\u002Fblock`, etc.)\n- world max height for newly created worlds (assumes tallest point is 10000 real-world meters)\n- 2 is recommended for a good balance of scale and playability. Use 1 for smaller, more compressed worlds.\n- Lower values put more stress on the GPU (Terrain Diffusion runs more often), while higher values put more stress on the CPU (larger world height). Most modern GPUs will be bottlenecked by the CPU around scale 2 or 3.\n\n## Common Issues\n\n**A dynamic link library (DLL) initialization routine failed**\n\nThis can happen for some older Java versions. Please update to the most recent version of Java 21 or higher. The [latest Microsoft OpenJDK 21](https:\u002F\u002Flearn.microsoft.com\u002Fen-us\u002Fjava\u002Fopenjdk\u002Fdownload) version is known to work.\n\n**LoadLibrary failed with error 126** *(CUDA build only)*\n\nThis is typically due to an improper CUDA or cuDNN installation. See [CUDA_INSTALL.md](CUDA_INSTALL.md) for troubleshooting steps.\n\n**java.lang.IllegalStateException: Failed to load terrain-diffusion models**\n\nThis typically indicates an \"out of memory\" error (the logs should show this as well).\nTerrain Diffusion's models take up about 2.5GB of RAM, so make sure to allocate enough RAM to account for this.\n\n**If your issue is still not resolved, please [raise it here](https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion-mc\u002Fissues\u002Fnew).**\n\n## Building from Source\n\nAn internet connection is required during the build to fetch the pinned model manifest metadata from Hugging Face.\n\nThe `-windows` build requires `libs\u002Fonnxruntime-dml.jar`, which is provided as part of the repo. See [Building onnxruntime with DirectML](#building-onnxruntime-with-directml) to build from source. \n\nBuild for Windows (DirectML):\n```\n.\u002Fgradlew build -PuseDml=true\n```\n\nBuild for CUDA:\n```\n.\u002Fgradlew build -PuseCuda=true\n```\n\nBuild for CPU (also handles macOS\u002FCoreML automatically):\n```\n.\u002Fgradlew build -PuseCpu=true\n```\n\nBuild all:\n```\n.\u002Fgradlew buildAll\n```\n\n### Building onnxruntime with DirectML\n\n**Requirements**\n\n- [Windows 10 SDK (10.0.17134.0)](https:\u002F\u002Fdeveloper.microsoft.com\u002Fen-us\u002Fwindows\u002Fdownloads\u002Fsdk-archive\u002Findex-legacy) — for Windows 10 version 1803 or newer\n- Visual Studio 2017 toolchain — install *Desktop development with C++* from the VS Installer\n- Visual Studio 2022 toolchain — same as above\n- Python 3.10+: [https:\u002F\u002Fpython.org\u002F](https:\u002F\u002Fpython.org\u002F)\n- CMake 3.28 or higher\n\nKeep both VS toolchains up to date. Full details at the [ONNX Runtime build docs](https:\u002F\u002Fonnxruntime.ai\u002Fdocs\u002Fbuild\u002Finferencing.html) and the [DirectML EP requirements](https:\u002F\u002Fonnxruntime.ai\u002Fdocs\u002Fexecution-providers\u002FDirectML-ExecutionProvider.html#build).\n\n**Steps**\n\nRun all commands from the **Developer Command Prompt for VS 2022**.\n\n```\ngit clone --recursive https:\u002F\u002Fgithub.com\u002FMicrosoft\u002Fonnxruntime.git\ncd onnxruntime\n.\\build.bat --config RelWithDebInfo --build_shared_lib --parallel --compile_no_warning_as_error --skip_submodule_sync --use_dml --build_java --build\n```\n\nThe built jar appears in `java\u002Fbuild\u002F`. Rename it to `onnxruntime-dml.jar` and place it in `libs\u002F` in this repository.\n\n## Note For Mod Developers\n\nWhile modifying the AI terrain itself is quite complex, the integration with Minecraft biomes is extremely simple. The model outputs elevation + 4 climate variables, and this is converted to Minecraft biomes with hand-written rules. This is the most immediate way to improve the quality of the terrain and is relatively easy, but takes time to get realistic. The entire biome classifier is [only 250 lines](https:\u002F\u002Fgithub.com\u002Fxandergos\u002Fterrain-diffusion-mc\u002Fblob\u002Fmaster\u002Fsrc\u002Fmain\u002Fjava\u002Fcom\u002Fgithub\u002Fxandergos\u002Fterraindiffusionmc\u002Fpipeline\u002FBiomeClassifier.java).\n\nThe terrain diversity far outpaces the biome diversity and there's a real opportunity to close that gap. I'm hoping someone goes crazy with it.\n","该项目是一个基于扩散模型的Minecraft地形生成模组。其核心功能是利用先进的AI技术自动生成多样化的地形，支持Windows、CUDA和CPU三种构建版本，其中Windows版本推荐使用现代GPU以获得最佳性能，而Mac用户则可享受CoreML带来的硬件加速。该模组适用于想要在Minecraft中探索由AI创造的独特自然景观的玩家或开发者。它要求安装Fabric加载器及相应的API，并且建议至少有1.5GB显存和2.5GB内存。通过内置的Web UI，用户可以轻松地探索生成的世界并定位感兴趣的地理特征。",2,"2026-06-11 03:55:08","trending"]