# xandergos/terrain-diffusion

Procedural generation with diffusion models (SIGGRAPH '26)

Repository: https://github.com/xandergos/terrain-diffusion
Canonical: https://ross.abutalabs.com/products/terrain-diffusion
Homepage: https://xandergos.github.io/terrain-diffusion/
Language: Python
License: MIT
License Family: permissive
Last push: 2026-08-12T03:58:54+00:00

## Health v2 (maintenance only)
Score: 64/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 97, release rhythm 29, longevity 49
- inputs: {"age_days": 694, "days_push": 21, "days_rel": 260, "gap_med": null, "n_releases_24m": 1}
- flags: prerelease_only
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1365, forks 85 (observed 2026-08-28T04:04:31.165437+00:00)

## What it is
Terrain Diffusion is a Python framework that uses diffusion models as a learned, deterministic replacement for Perlin noise, generating infinite, randomly-accessible planetary terrain heightmaps in real time. Its core InfiniteDiffusion algorithm makes diffusion sampling lazy and unbounded with O(1) random access and seed-consistency, built on the infinite-tensor library.

## Use cases
- generate infinite realistic terrain heightmaps for games
- replace Perlin noise with learned procedural terrain generation
- stream planetary-scale terrain and climate data in real time
- generate infinite panoramas with Stable Diffusion deterministically
- convert hand-drawn coarse maps into high-resolution heightmaps
- build Minecraft-style world generators with diffusion models

## When to choose
- you need infinite, seed-consistent, randomly-accessible terrain generation with learned realism
- you want diffusion-model quality output at interactive rates on a consumer GPU
- you need a drop-in procedural-noise-like interface backed by neural models
- you want to study or reproduce the InfiniteDiffusion algorithm from SIGGRAPH '26

## When to avoid
- you need a lightweight CPU-only solution with fast generation
- you only need simple classical noise like Perlin or simplex
- you need a stable production PyPI package (the published package is a 0.0.1 placeholder)
- you work on Mac and need GPU acceleration (Mac is CPU-only)

## Facets
- artifact type: library
- maturity: active
- function: machine-learning, deep-learning, llm-inference, simulation, graphics, game-engine
- domain: machine-learning, deep-learning, graphics, simulation
- platform: python, windows, cross-platform
- tags: diffusion-models, procedural-generation, terrain-generation, infinite-generation, heightmaps, pytorch, research-paper, siggraph, game-development, gpu, linux, macos

## Member repositories
- xandergos/terrain-diffusion (main) score 64

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:04:31.165437+00:00.
- Health v2: computed from the inputs above; adoption is never an input.
- Inferred fields (summary, facets, guidance): AI-extracted, prompt v1, taxonomy v1, on 2026-08-30T04:41:17.493402+00:00, confidence not recorded.
  - readme: https://github.com/xandergos/terrain-diffusion (fetched 2026-08-28T04:04:31.165437+00:00, sha 581059e6b5d6)
  - homepage: https://xandergos.github.io/terrain-diffusion/ (fetched 2026-08-29T11:58:36.393555+00:00, sha 95798b54f803)
  - registry_pypi: https://pypi.org/pypi/terrain-diffusion/json (fetched 2026-08-29T11:58:36.402645+00:00, sha 991691f9a710)
- Data as of 2026-08-30T08:39:29.467469+00:00.