---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: file_name
    dtype: string
  - name: type
    dtype: string
  - name: project_id
    dtype: string
  - name: mod_slug
    dtype: string
  - name: author
    dtype: string
  - name: license
    dtype: string
  - name: project_type
    dtype: string
  - name: version_url
    dtype: string
  - name: text
    dtype: string
  splits:
  - name: train
    num_bytes: 733001168.875
    num_examples: 1034057
  download_size: 387158319
  dataset_size: 733001168.875
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
license: mit
size_categories:
- 1M<n<10M
tags:
- minecraft
- pixelart
- pixel
- texture
- 16x16
- mc
---

# Minecraft 16x Text-to-Image Dataset (Captioned)

## Description
This dataset contains over **1 million Minecraft textures** in 16x16 resolution. It has been specifically processed and captioned for training generative AI models (Text-to-Image). 

Each image is paired with a descriptive natural language caption derived from the original file labels, enabling AI models to learn the relationship between Minecraft concepts (blocks, items, tools) and their pixel-art representation.

## Data Source & Attribution
This is an enhanced and captioned version of the following source dataset:
* **Original Dataset:** [OVAWARE/16xModdedMinecraft](https://huggingface.co/datasets/OVAWARE/16xModdedMinecraft)
* **Content:** Textures gathered from a wide variety of Minecraft mods in 16x16 resolution.

## Processing Details
1. **Natural Language Captioning:** Raw labels were cleaned and transformed into descriptive prompts (e.g., `gold_ingot` -> `a 16x16 pixel art minecraft texture of gold ingot, high quality game sprite`).
2. **Optimization:** Formatted for seamless compatibility with Hugging Face `diffusers` and other training scripts (Stable Diffusion, LoRA, etc.).

## Intended Use
This dataset is designed for:
- Training generative models for texture creation.
- Fine-tuning (LoRA) to inject "Minecraft style" into larger models.
- Research on low-resolution constrained pixel art generation.

## License
This dataset is distributed under the **MIT License**, following the license of the original source data.