Watch the skill in action — generating a complete 64×64 character spritesheet from a single prompt:
View on X (Twitter)
For: Codex (OpenAI) and GPT Web Agent
Purpose: Generate game-ready 64×64 pixel-art character spritesheets with 8-direction animations from a text prompt or reference image.
"Use game-character-64 to make a 64x64 knight with 8-direction walk and attack
animations from this image."
Given a concept or reference image, this skill:
- Establishes character identity — species, class, silhouette, palette, weapon, outfit.
- Generates a canonical base sprite — a readable 64×64 pixel-art character.
- Produces animation strips — idle, walk, attack for all 8 directions (24 rows × 6 columns).
- Runs pixel cleanup — quantizes palettes, snaps edges, handles alpha.
- Validates the atlas — checks frame integrity and generates a contact sheet.
- Packages output — final spritesheet + QA report ready for import into RPGs, roguelikes, top-down action games, or tactics engines.
run/
├── prompts/
├── generated/ # raw strips from image generation
├── final/
│ ├── character-sheet.png
│ └── character-sheet-clean.png
└── qa/
├── validation.json
└── contact-sheet.png
Default 8-direction order:
south, south-east, east, north-east, north, north-west, west, south-west
| Action | Frames | Notes |
|---|---|---|
| idle | 4 | padded to 6 cells |
| walk | 6 | full cycle |
| attack | 6 | full cycle |
Default atlas: 384 × 1536 (6 columns × 24 rows × 64px).
Smaller scopes available:
walk-only: 8 rows × 6 framesidle-walk: 16 rows × 6 framescombat: idle + walk + attack for all 8 directions
"Use game-character-64 to make a 64x64 knight with 8-direction walk and attack
animations from this image."
scripts/pixel_snap.py— alpha threshold, palette quantization, pixelate-scale edge hardeningscripts/validate_64_sheet.py— frame validation + contact-sheet generationscripts/export_animation_previews.py— export GIF/PNG previews per animation row
- Python 3.9+
- Pillow (
pip install Pillow) - Image generation backend (Codex/GPT
$imagegenor DALL-E / Midjourney / etc.)
MIT — see SKILL.md for full workflow documentation.