# AvaLovelace1/BrickGPT

[ICCV 2025 Best Paper] Official repository for BrickGPT, the first approach for generating physically stable toy brick models from text prompts.

Repository: https://github.com/AvaLovelace1/BrickGPT
Canonical: https://ross.abutalabs.com/products/brickgpt
Homepage: https://avalovelace1.github.io/BrickGPT/
Language: Python
License: MIT
License Family: permissive
Topics: generative-ai, lego
Last push: 2026-05-21T22:17:38+00:00

## Health v2 (maintenance only)
Score: 57/100 (v2, computed 2026-09-03T02:20:16.233290+00:00)
- activity 83, release rhythm 35, longevity 37
- inputs: {"age_days": 529, "days_push": 104, "days_rel": null, "gap_med": null, "n_releases_24m": 0}
- flags: no_releases
- formula: round(0.45*activity + 0.35*rhythm + 0.20*longevity); archived -> min(score, 10)

## Adoption (not part of the score)
Stars 1701, forks 113 (observed 2026-08-28T04:05:24.315297+00:00)

## What it is
BrickGPT is the official implementation of an ICCV 2025 Best Paper approach that generates physically stable, buildable toy brick (LEGO) models from text prompts using a fine-tuned autoregressive LLM with physics-aware rollback. It includes the StableText2Brick dataset of 47,000+ brick structures and supports human or robotic assembly of the generated designs.

## Use cases
- generate lego brick models from text prompts
- train a language model to predict brick structures
- check physical stability of brick assemblies
- convert 3d meshes into buildable brick layouts
- research on physically grounded generative models
- robotic assembly of generated brick designs

## When to choose
- you want to generate stable, buildable brick structures from natural language
- you need the StableText2Brick dataset for research on structured 3D generation
- you are studying physics-aware constrained decoding for LLMs
- you want reproducible code from a peer-reviewed generative AI paper

## When to avoid
- you need general-purpose 3D mesh generation rather than brick/voxel structures
- you cannot obtain a Llama-3.2 gated model access token or Gurobi license
- you need a production-ready end-user application rather than research code
- you want photorealistic 3D assets instead of toy brick models

## Facets
- artifact type: library
- maturity: active
- function: llm-training, llm-inference, machine-learning, data-generation, simulation
- domain: artificial-intelligence, large-language-models, deep-learning, machine-learning, graphics, gaming-tools
- platform: python, cross-platform
- tags: lego, text-to-3d, generative-ai, physics-stability, iccv-2025, research-code, stabletext2brick, gpu

## Member repositories
- AvaLovelace1/BrickGPT (main) score 57

## Provenance
- Observed fields: from GitHub, fetched 2026-08-28T04:05:24.315297+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 2026-08-30T03:37:41.727630+00:00, confidence not recorded.
  - readme: https://github.com/AvaLovelace1/BrickGPT (fetched 2026-08-28T04:05:24.315297+00:00, sha e3a22d14b3f6)
  - homepage: https://avalovelace1.github.io/BrickGPT/ (fetched 2026-08-29T11:12:09.380510+00:00, sha 8eee72abe13a)
- Data as of 2026-08-30T08:39:29.467469+00:00.