Create pixel art with Retro Diffusion models on Replicate
Retro Diffusion has crafted a beautiful set of models and styles for creating game assets, character sprites, tiles, and other wonderful retro graphics. These models are specifically trained for grid-aligned, limited-palette pixel art. Now you can run them on Replicate.
There are four models available on Replicate from Retro Diffusion: rd-fast, rd-plus, rd-tile, and rd-animation. Each of these models offers numerous style presets and supports arbitrary width/height, input images, palette images, background removal, and seamless tiling.
The rd-fast model is optimized for fast pixel art image generation and supports 15 styles, from portraits to Minecraft-style items. For example, you can use the following command in Python to create an image: output = replicate.run( "retro-diffusion/rd-fast", input={ "prompt": "knight character, retro game asset, side view", "style": "game_asset", "width": 128, "height": 128, "num_images": 1, "remove_bg": True, }, ).
The rd-plus model provides high quality and authentic pixel art image generation. It offers styles for scenes, maps, UI, and icons. To create a JRPG-style map, you can use the following code: output = replicate.run( "retro-diffusion/rd-plus", input={ "prompt": "top-down forest path with trees and rocks, classic jrpg map", "style": "topdown_map", "width": 256, "height": 256, "num_images": 1, "tile_x": True, "tile_y": True, }, ).
The rd-tile model provides all the tools you need for generating pixel art tilesets. It focuses on single tiles and their variations. An example command for creating a tileset is: output = replicate.run( "retro-diffusion/rd-tile", input={ "prompt": "grass and dirt path tileset with stones and flowers", "style": "tileset", "width": 32, "height": 32, "num_images": 1, }, ).
The rd-animation model generates animated sprites or spritesheets with consistent framing and low frame counts that match common game engines. To create a character animation, you can use: output = replicate.run( "retro-diffusion/rd-animation", input={ "prompt": "small pirate character with parrot, walking animation", "style": "four_angle_walking", "width": 48, "height": 48, "return_spritesheet": True, }, ).
These examples use Python, but you can also call the models from JavaScript and other languages with Replicate’s SDKs. Kudos to the Retro Diffusion team, including Atropulse and nerijs, for creating these amazing tools.
Explore the capabilities of Seedream 5.0 for image creation
Extract Text from Documents and Images with Datalab Marker and OCR
Похожие статьи
How AI Becomes the First Analyst on Your Team
AI is becoming the first analyst in teams, changing approaches to data analysis and decision-making.
Revise Governance to Achieve AI Goals
Agentic AI changes the approach to governance and security in organizations.
Explore Together AI Innovations at NVIDIA GTC 2026
Together AI showcases innovations at NVIDIA GTC 2026, including new models and capabilities.