moss-pixelart-v.0.3

Model Description

moss-pixelart-v.0.3 is a Low-Rank Adaptation (LoRA) model fine-tuned on OpenMOSS-Team/MOSS-Image-SFT specifically for generating clean, grid-aligned pixel art and 2D game assets.

The model is trained to render at a strict 8x pixel scale, where every 8×8 block of output pixels represents a single logical pixel. This enforces strong grid adherence during generation and makes downsampling to native pixel resolutions straightforward.

Baked with <3 at fingergun.dev

Resolution Mapping (8:1 Scale)

Generation Resolution    Native Pixel Art Grid   Downscale Factor
512 × 512 (Training res) 64 × 64 px              8x (12.5%)
768 × 768                96 × 96 px              8x (12.5%)
1024 × 1024              128 × 128 px            8x (12.5%)

Usage & Inference

Trigger Word

Prepend pxl8x (or [pxl8x]) to your prompt to activate the 8x pixel art style.

Recommended Inference Settings (MOSS-Image Defaults)

This LoRA is designed to run using the standard MOSS-Image-SFT default parameters:

Sampler / Scheduler: Euler (flowmatch / sgm_uniform)
Inference Steps: 50 (or 30–50 steps for standard generation)
Guidance Scale (CFG): 3.0 (recommended range 3.0–6.0)
Flow Shift: 5.0
Skip Layer Guidance (SLG): Unconditional skip on layer 22 (uncond_skip_layers: [22], slg_start: 0.01, slg_stop: 1.0)
LoRA Weight: 1.0

Turbo / Step-Distilled LoRA Compatibility

moss-pixelart-v.0.3 stacks cleanly with MOSS Turbo / DMD (8-step) distillation LoRAs. When pairing with an 8-step Turbo LoRA:

Set Inference Steps to 8.
Keep Sampler on Euler and Flow Shift at 5.0.
Keep both LoRA weights around 1.0.

Example Prompts

pxl8x an isometric cyberpunk ramen stall at night, glowing neon sign reading "NOODLES", steam rising from bowls, rain puddles on asphalt, crisp pixel art
pxl8x a retro sci-fi space marine in bulky industrial armor holding a motion tracker, dark derelict spacecraft corridor, green CRT monitor glow
pxl8x a 3/4 top-down RPG potion shop interior, wooden shelves packed with glowing glass vials, a sleepy alchemist cat on the counter, warm lantern light
pxl8x a classic 16-bit boss sprite of a mechanical dragon breathing plasma fire, dark volcanic cavern background, high contrast arcade palette

Post-Processing (Recommended)

While the model maintains an 8×8 pixel grid natively, diffusion decoding can occasionally introduce subtle sub-pixel edge bleeding or slight grid drift ("mixels").

For game-ready 1:1 assets, post-process your raw outputs with Sprite Fusion Pixel Snapper (GitHub). It automatically detects the pixel grid, cleans up blurry or inconsistent blocks, and snaps the image down to a crisp, true-resolution sprite sheet or tile.

Training Details & Hyperparameters

Trained via Ostris's ai-toolkit (diffusion_trainer) using the following configuration:

Architecture & Network

Base Model: OpenMOSS-Team/MOSS-Image-SFT (moss_image arch)
Network Type: LoRA (diffusers format, bf16)
Linear Rank / Alpha: 32 / 32
Conv Rank / Alpha: 16 / 16
Excluded Layers: layers.22. (paired with uncond_skip_layers: [22] on the base model)
Base Quantization: convrot8 (applied to both Transformer and Text Encoder)
Trained Components: Transformer / UNet only (train_text_encoder: false)

Training Configuration

Dataset: pixel snapped 64x64 images (512x512 resolution, flip_x: false, flip_y: false)
Training Steps: 3000
Batch Size: 2 (Gradient Accumulation 2 = Effective Batch Size 4)
Optimizer: adamw8bit (lr: 0.0001, weight_decay: 0.001)
Noise Scheduler: flowmatch (timestep_type: shift)
Loss Type: mse
Differential Guidance: Enabled (scale: 3.0)
Caption Dropout Rate: 0.05 (5%)
Precision: bf16

Limitations

Resolution Multiples: Always generate at dimensions divisible by 16 (and ideally multiples of 64 like 512x512, 768x768, or 1024x1024) to keep the 8x pixel grid evenly aligned with the VAE latent space.
Fine Text At Low Resolutions: Generating at 512x512 yields a 64x64 logical canvas—keep in-image text short (1–2 words in block font) or generate at 1024x1024 (128x128 logical grid) if you need readable signage.