šÆ What does this model do?
This LoRA transforms any single object image into a professional 2Ć2 sprite sheet with 4 different camera angles:
āāāāāāāāāāāāāāāāāāāāāā¬āāāāāāāāāāāāāāāāāāāāā
ā ISOMETRIC ā ā ISOMETRIC ā ā
ā Front-right view ā Front-left view ā
āāāāāāāāāāāāāāāāāāāāāā¼āāāāāāāāāāāāāāāāāāāāā¤
ā SIDE VIEW ā ā TOP-DOWN ā ā
ā Profile from left ā Bird's eye view ā
āāāāāāāāāāāāāāāāāāāāāā“āāāāāāāāāāāāāāāāāāāāā
Perfect for game developers, asset creators, and 3D artists who need consistent multi-view references.
š¼ļø Examples
| Input | Output (2Ć2 Sprite Sheet) |
|---|
 |  |
 |  |
 |  |
 |  |
š Quick Start
| Parameter | Value |
|---|
| Prompt | 2x2 sprite sheet |
| Recommended LoRA Scale | 1.1 |
š» Usage
Try it Live on fal.ai
ā¶ļø Open Playground
With fal.ai SDK
import fal_client
def on_queue_update(update):
if isinstance(update, fal_client.InProgress):
for log in update.logs:
print(log["message"])
result = fal_client.subscribe(
"fal-ai/flux-2-klein/4b/base/edit/lora",
arguments={
"prompt": "2x2 sprite sheet",
"model_name": None,
"loras": [{
"path": "https://huggingface.co/fal/flux-2-klein-4b-spritesheet-lora/resolve/main/flux-spritesheet-lora.safetensors",
"scale": 1.1
}],
"embeddings": [],
"image_urls": ["https://your-image-url.png"]
},
with_logs=True,
on_queue_update=on_queue_update,
)
print(result)
š Output Layout
The model produces a 2Ć2 grid with these specific views:
| Position | View Type | Direction |
|---|
| Top-Left | Isometric 3/4 | Object faces bottom-right ā |
| Top-Right | Isometric 3/4 | Object faces bottom-left ā |
| Bottom-Left | Side Profile | Object faces left ā |
| Bottom-Right | Top-Down | Object faces up ā |
All 4 views maintain the same art style, colors, and proportions.
šØ About the Red Background
The output uses a consistent red background by design. This solid color background:
- Improves style consistency across all 4 views
- Makes it easy to remove the background for game integration
- Ensures clean edges for sprite extraction
š¦ Model Files
| File | Use Case |
|---|
flux-spritesheet-lora.safetensors | fal.ai |
mFtSubaAlDqtHyKk6eIPi_pytorch_lora_weights_comfy_converted.safetensors | ComfyUI |
š Training Details
Click to expand
Dataset
- Size: 48 curated sprite sheet pairs (from 120 generated, manually validated)
- Content: Diverse vehicle and object categories including:
- Aircraft: jet fighters, helicopters, biplanes, cargo planes, drones, zeppelins
- Military Vehicles: tanks, APCs, artillery, humvees, missile launchers
- Civilian Vehicles: sports cars, motorcycles, trucks, buses, bicycles
- Construction Equipment: excavators, bulldozers, cranes, forklifts
- Trains: locomotives, bullet trains, freight trains, trams
- Ships & Boats: battleships, submarines, yachts, speedboats, kayaks
- Space & Sci-Fi: rockets, space shuttles, mechs, hover vehicles, UFOs
- Style: Game-ready 2D assets with red background for consistency
- Validation: AI-validated for correct view orientations, then manually curated
Training
- Base Model: FLUX.2-Klein 4B
- Platform: fal.ai
- Method: LoRA training
š® Use Cases
- Game Development: Generate sprite sheets for 2D/isometric games
- Asset Creation: Quick multi-view references for 3D modeling
ā ļø Limitations
- Output background is solid red (by design, for consistency)
š License
Apache 2.0