pixel_assets_general_ernie_v1

Prompt

A pixel art sprite of a medieval knight wearing metal armor and a helmet with a red plume, a sword in one hand and a shield in the other hand. The background is white.

Prompt

A pixel art image of a witch with long red hair and blue eyes, wearing a purple hat and robes trimmed with white and light purple colors. White background

Prompt

A pixel art image of a man with light brown hair in a long ponytail. He is wearing a butler outfit and leaning forward towards the viewer holding a bowl of soup. The background is a fancy restaurant with dining tables in the back, a chandelier, and a painting of a noblewoman on the left walls.

Prompt

A pixel art aerial shot of an island in the middle of the ocean. On the right side of the island is a giant metal orb building with a satellite on top of it.

Prompt

A pixel art scene of wide, vast sand dunes with a tall cylindrical tower in the far background surrounded by a sandstorm. In the foreground is a medieval carriage being pulled by a triceratops moving towards the tower.

Prompt

A pixel art image of a top-down view of stairs leading down into darkness. In the background the darkness has several faint red eyes.

Prompt

A pixel art sprite of a short red dagger with a green poisoned tip on the top right and a gray and brown hilt. There are 16 copies of the sprite in a 4 by 4 grid. The background is white.

Prompt

A pixel art sprite of a rectangular orange shield with the head of a gray dragon with its mouth open facing straight. The background is white.

Prompt

A pixel art sprite of a skeleton warrior wearing a helmet with two horns, holding an axe with both hands raised, facing right, white background

Model description

A pixel art LoRA for general-purpose game assets such as character sprites, creatures, items/equipment, backgrounds, scenery, and icons.

How to use

You can use the default ERNIE-Image-Turbo workflow from ComfyUI, and no prompt enhancer needed. The sample images also have workflows.

How to get pixel-perfect images

Downscale by a factor of 4. So 512x512 images should downscale to 128x128, 1024x1024 to 256x256, and so on. Using k-centroid with something like PixelOE works well.

See examples below:

Raw output	K-centroid downscaled, then upscaled back 4x

Does this LoRA work with ERNIE-Image base?

Yes, but I don't recommend it. The LoRA is meant to be used with the turbo model. For some reason, outputs with the base model are very bad. The colors are way too bright or saturated, and there are more issues with anatomy. Maybe there's a problem with my settings.

Notes & Issues

There are still some issues with certain prompts with the ERNIE turbo model.

• The model tends to make characters face forward or in a 3/4 angle even if your prompt has a different view. This might just be a limit of the turbo model, though.
• If prompting for sprites, make sure to include "white background" somewhere, otherwise you'll sometimes get a detailed background.
• Since I trained this on a 4x upscaled pixel art dataset, if you want smaller sprites, just prompt for copies of a sprite in a 2x2 or 4x4 grid (see the sample images).
• The dataset this LoRA was trained on contains 512x512, 768x768, and 1024x1024 images, but you can change the resolution and still get decent images.

Download model

Download them in the Files & versions tab.