Skip to content
Images

LiliPixel-V1

A compact model for pixel art and sprite generation

2 min readupdated

LiliPixel-V1 is a diffusion transformer model (DiT) with 191 million parameters, trained from scratch to generate pixel art and sprites at native resolutions of 16x16, 32x32, and 64x64 pixels. The model uses the Voyage-4-nano text encoder and supports transparent backgrounds for low resolutions, as well as a realistic graphics mode for 64x64. Thanks to its modest size, it requires only about 1 GB of VRAM to run in BF16 format. The tool is well-suited for game developers and pixel art enthusiasts for quick asset generation.

Key features

  • Trained from scratch on pixel art datasets with support for 16x16, 32x32, and 64x64 resolutions
  • Low system requirements: only about 1 GB of VRAM for fast generation
  • Support for transparent backgrounds (RGBA) for 16x16 and 32x32 resolutions
  • Built-in photorealism mode for 64x64 resolutions using a special prompt

What you can do

  • Generate character sprites and item icons with transparent backgrounds for indie games
  • Create small pixel art landscapes and scenes at 64x64 resolution
  • Quickly experiment with prompts to get collections of textures and icons

Example

import torch
from diffusers import DiffusionPipeline

pipe = DiffusionPipeline.from_pretrained(
    "Pinkstack/LiliPixel-V1-190M",
    custom_pipeline="Pinkstack/LiliPixel-V1-190M",
    trust_remote_code=True,
    torch_dtype=torch.bfloat16,
).to("cuda")

image = pipe(
    prompt="cute border collie dog with a purple colar, transparent background",
    height=32, width=32,
    num_inference_steps=30,
    guidance_scale=3.0,
    generator=torch.Generator("cuda").manual_seed(96),
).images[0]
image.save("LiliPixel-image.png")
Read moreDetails
Model: 191M parameters, DiT architecture (24 layers, width 768, 16 heads)
Requirements: ~1 GB VRAM, BF16 inference support
Training: 132,000 steps on a single RTX 4080 (126 GPU hours)
Limitations: complex scenes and a large number of objects in low resolution may turn out unclear, generating multiple variations is recommended
Discuss Telegram 35 141

Will it run?

Runs in F32

Fits without quantisation.

  • F32weights 0.7 GB · ready files2.3 GB
  • BF16weights 0.4 GB1.9 GB
  • FP8 / INT8recommendedweights 0.2 GB1.7 GB
  • 4-bitweights 0.1 GB1.6 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 · F32 — 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.

    # Pinkstack/LiliPixel-V1-190M
    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 Pinkstack/LiliPixel-V1-190M --local-dir ~/models/LiliPixel-V1-190M
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to LiliPixel-V1-190M
  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 LiliPixel-V1-190M (quants, finetunes, LoRAs) will show up under the bell.

No known relatives of this model yet. All derivatives on Hugging Face

About the model

Pinkstack/LiliPixel-V1-190M
Author
Pinkstack
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

14135

Ask AI about this model

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

Comments