Dataset · Angais
1.4m pixel art sprites with captions
Angais/pixel-art-atlas-1.4m
Synthetic 16x16 game sprites with English captions for training text-to-image models and LoRA.
What's inside
The dataset contains 1,409,404 synthetic game sprites covering 6,194 training subjects. Each image is a native 16x16 sprite with a transparent background and an English caption. The sprites were generated from text prompts using Krea 2 Turbo and pixel art LoRA; image-caption alignment has not been systematically verified.
What you can do with it
- Train text-to-image LoRA for pixel art
- Generate game assets
- Evaluate models on 16x16 sprites
- Tasks
- Text → image
- Languages
- English
- Size
- 1M–10M rows
- Downloads this month
- 75.3K
- Likes
- 675
Models trained on it
- tomaarsen/TinyCLIP-39M-PixelArtSentence similarity 897
- unsupervised-abdullah/pixel-art-sdxl-loraText → image 222
- Pinkstack/LiliPixel-V1-190MText → image 216
- ihor-umanets/pixelsprite-canvas-v2Text → image 195
- MahdiAlikhah/pixelquill-61mText → image 96
- EmilioSchwaiger/pixie-16Text → image 73
- DexagonAI/indepth-10MText → image 37
- ihor-umanets/pixelsprite-canvas-v3Text → image 27
How to load it
from datasets import load_dataset
ds = load_dataset("Angais/pixel-art-atlas-1.4m")