Skip to content
Images

pixelcbe-klein4b-lora

pixel-art city rendering for FLUX.2

2 min readupdated

This edit-LoRA (a lightweight add-on for image editing) turns flat isometric 3D city renders into detailed pixel art in the style of pixelcbe.com. Built by codeanand on top of the compact black-forest-labs/FLUX.2-klein-4B, it is designed specifically for image-to-image workflows. Unlike standard filters or text-to-image generators, the adapter was trained strictly on pairs of base renders generated from OpenStreetMap data and hand-drawn pixel art from the exact same perspective, so it faithfully preserves both geography and camera angle. It is a great fit for game developers, cartographers, and procedural isometric map enthusiasts.

Key features

  • Stylizes flat isometric base renders (1024×1024) into hand-crafted pixel art while preserving the original color palette and camera angle.
  • Fast inference — generating a single tile takes about 1.5 seconds in 4 steps on an A100 GPU.
  • Performs best on dense urban layouts typical of Indian cities.
  • Built to work alongside the open-source pixelcbe pipeline, where terrain, stadiums, airports, and overpasses are handled procedurally in separate passes.

What you can do

  • Restyle isometric 3D blockouts of real-world neighborhoods built from OpenStreetMap data into cozy pixel-art maps.
  • Pair it with the base FLUX.2-klein-4B model and the render_v2.py script to quickly generate map tiles for city builders or web maps.

Example

Feed in a 1024×1024 flat isometric render of a city block generated by base/render_v2.py (with a locked camera angle and palette). In 4 inference steps, the model outputs the exact same map tile, complete with pixel-art textures and fine details.

Read moreDetails
Base model: black-forest-labs/FLUX.2-klein-4B (Rank-64 edit-LoRA).
Training: 2,000 steps using ostris/ai-toolkit on a single A100 GPU (~$5 budget).
Inference: 4 steps, ~1.5s per 1024×1024 tile on an A100.
Limitations: best suited for dense Indian urban layouts; usage is governed by the base model's license.
Discuss Telegram 1

Will it run?

Runs in BF16

Fits without quantisation.

  • BF16weights 7.4 GB · ready files9.6 GB
  • FP8 / INT8recommendedweights 3.8 GB5.7 GB
  • 4-bitweights 2.2 GB3.9 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 3080

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

This is a LoRA: counted together with its base model FLUX.2-klein-4B.

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.

    # codeanand/pixelcbe-klein4b-lora
    # ===== Tokens: paste yours between the quotes =====
    export HF_TOKEN=""        # huggingface.co/settings/tokens, for gated models
    # ==========================================
    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 black-forest-labs/FLUX.2-klein-4B --local-dir ~/models/FLUX.2-klein-4B
    hf download codeanand/pixelcbe-klein4b-lora --local-dir ~/models/pixelcbe-klein4b-lora
    # The model is in ~/models. How to run it is in the model card (link above).
    Add more LoRAs to FLUX.2-klein-4B
  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

Requires the base model

black-forest-labs/FLUX.2-klein-4B

This model is a LoRA / adapter. Download the base model first, then load this one on top.

New posts about versions of FLUX.2-klein-4B (quants, finetunes, LoRAs) will show up under the bell.

Quantised versions 12All on Hugging Face
Fine-tunes 12All on Hugging Face
LoRAs and adapters 12All on Hugging Face

About the model

codeanand/pixelcbe-klein4b-lora
Author
codeanand
Category
Images
Released
Posted

On NeyroLearn

Hugging Face

01

Ask AI about this model

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

Comments

Other

Iris-Z

Iris-Z