Skip to content
Images

UbaNova-32

pixel art sprite generator

2 min readupdated

UbaNova-32 is a compact 42.8M-parameter autoregressive model created by Muhammad Fahril Irkham to generate 32x32 2D pixel art from English text prompts. It predicts pixels sequentially in raster order, guided by a frozen CLIP (computer vision and text processing model) text encoder alongside control tokens for size, background, canvas fill ratio, and detail density. The model outputs crisp sprites with exact palette colors and transparent backgrounds—no blurring or post-processing required—making it a handy tool for indie game devs and content creators.

Key features

  • Sprite generation: creates detailed 32x32 pixel art with a transparent background from text prompts.
  • Fine-grained control: lets you adjust output size (including a 16x16 mode), background, canvas fill ratio, and color density.
  • Autoregressive architecture: a 42.8M-parameter decoder-only transformer that predicts pixels one by one in raster order.
  • Lightweight and efficient: runs on a GPU with FP16 autocast or on a standard CPU, with weights taking up just 86 MB.

What you can do

  • Build game assets: generate sprite sets for 2D games, including weapons, potions, characters, and environment props.
  • Design UI elements: produce clean pixel icons for retro projects and indie titles.
  • Experiment with styles: tweak prompts and detail settings to get classic 16x16 or 32x32 retro graphics.

Example

from infer32 import SpriteGenerator32

gen = SpriteGenerator32()                 # cuda если доступно, иначе cpu
images = gen.generate("golden crown with rubies", n=4, seed=42)   # список 32x32 RGBA PIL изображений
images[0].resize((256, 256), 0).save("crown.png")
Read moreDetails
Architecture: decoder-only transformer (512 hidden width, 12 layers, 8 attention heads)
Parameters: 42.84M trainable parameters
Benchmarks: CLIP score 28.87, prompt retrieval accuracy R@1 50.0%
Weights: FP16, 86 MB, initialized from UbaNova-16
Examples
Discuss Telegram 2

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 0.1 GB · ready files1.6 GB
  • FP8 / INT8recommendedweights 0.0 GB1.5 GB
  • 4-bitweights 0.0 GB1.5 GB

Tip: go 8-bit.FP8 / INT8 is within about a percent of full precision in quality, takes half the memory and runs faster on RTX 40/50 and H100. Full precision only pays off for fine-tuning.

What to get for this model

Massed ComputeOpen Massed Compute

Your own GPU: FP8 / INT8 — from RTX 2060 · BF16 — from RTX 2060

Massed Compute hourly prices, November 2026. The discounted price applies with the coupon at checkout.

Estimate: weights plus working memory. Long videos, high resolutions and long contexts need more.

Run it on Massed Compute

The script downloads the model to the rented machine. How to run it is in the model card.

  1. 1. Rent a machine with a fitting GPU

    Open Massed Compute
  2. 2. Paste the script into the machine's terminal

    Connect over SSH (the Massed Compute panel shows the command) or open a terminal on the machine's desktop.

    HF_TOKEN / CIVITAI_TOKENThe script starts with token lines: paste yours between the quotes. The Hugging Face token is for gated models, the Civitai one for Civitai models.

    # ubanova/UbaNova-32
    mkdir -p ~/models
    curl -LsSf https://astral.sh/uv/install.sh | sh
    source "$HOME/.local/bin/env"
    uv tool install "huggingface_hub[cli]"
    hf download ubanova/UbaNova-32 --local-dir ~/models/UbaNova-32
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to UbaNova-32
  3. 3. Use it

    The model is in ~/models. How to run it is in the model card (link above).

Stop the machine in the panel when you are done: it bills by the minute while it runs.

How to speed it up

Picked for this model type. Choose your hardware above to see what fits you exactly.

  • SageAttention 3v2.2.0from October 28, 2025

    even faster on RTX 50 / Blackwell

    FP4 attention for Blackwell. Faster than SageAttention 2, but the authors suggest v2 when accuracy matters.

    How to enable: Build from the sageattention3_blackwell folder of the repo; in ComfyUI via KJNodes.

    Earlier generations (2)
    • SageAttention 2++ — Quantised (INT8/FP8) attention with no visible quality loss. The main speed-up for image and video generation.
    • SageAttention 1 — First version: INT8 for Q·K. Still works, but 2++ is faster and more accurate.
  • FlashAttention 44.0.0b40from November 18, 2026new release

    fastest exact attention on B200

    Kernels for Blackwell (beta). Exact attention without quantisation — fine for everything, LLMs included.

    How to enable: pip install flash-attn-4; vLLM and SGLang pick it up themselves.

    Earlier generations (3)
    • FlashAttention 3 — The Hopper (H100/H200) version with FP8.
    • FlashAttention 2 — The standard for RTX 30/40/50 and A100: exact attention without the memory cost of long contexts.
    • xFormers — The older way to speed up attention, works almost everywhere. Sage or Flash are faster where available.
  • cache-dit (DBCache)v1.5.6from November 20, 2026new release

    1.5–3× at nearly the same quality

    Reuses block outputs between diffusion steps. Supports most DiT models.

    How to enable: pip install cache-dit; one line cache_dit.enable_cache(pipe) for diffusers.

    Earlier generations (2)
    • MagCache — Skips steps by magnitude of change. More accurate than TeaCache, no calibration.
    • TeaCache — The first popular diffusion step cache. Works, newer methods are more accurate.
  • Nunchaku (SVDQuant 4-bit)v1.2.1from January 25, 2026

    ≈3× faster and 3.5× less memory

    4-bit versions of popular image models with almost no quality loss (FP4 on RTX 50).

    How to enable: Ready weights for supported models; in ComfyUI use ComfyUI-nunchaku.

  • torch.compile2.15.0from October 28, 2026new release

    10–30% for free

    Compiles the model for your GPU. Slower first run, faster afterwards.

    How to enable: model = torch.compile(model); in ComfyUI the TorchCompileModel node.

Model family

New posts about versions of UbaNova-16 (quants, finetunes, LoRAs) will show up under the bell.

Fine-tunes 1All on Hugging Face

About the model

ubanova/UbaNova-32
Author
ubanova
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

02

Ask AI about this model

Answered by Gemini 4 · the chat is kept only in this tab

Comments