Document radial overhead validation

Reject asymmetric radial overhead sprites

Reject incompatible overhead silhouettes

Strengthen strict overhead sprite prompting

Upgrade generation to full-quality SDXL

Enforce top-down sprite perspective

Improve generated image sharpness

Start generation form without sample prompts

Extract transparent assets with BiRefNet

Reject non-studio cutout attempts

Reject scene-contaminated transparent assets

Support fully custom asset manifests

Normalize generated sprite scale

Reject miniature duplicate sprites

Detect touching duplicate sprite subjects

Scale ZeroGPU duration by asset count

Keep generation prompts inside CLIP budget

Generate separate roles and reject multi-view sprite sheets

Switch to blue and yellow light theme

Redesign Space interface

Enforce single-subject neural assets

Document direct text-to-image generation

Generate assets directly from text prompts

Update README.md

Update app.py

Update README.md

Update app.py

Update README.md

Update app.py

Update requirements.txt

Fix procedural subject and background matching

Enforce camera perspective in background generation

Make validation aware of image model behavior

Add deterministic asset integration workflow

Start roles and theme blank

Add safe HF token validity check

Start with empty paste-friendly HTML input

Clean up forest background artifacts

Keep backgrounds separate and prevent player trails

Prevent missing images from black-screening canvas games

Use whole 2D sprite generator for sprite roles

Prevent sprite generations from becoming texture maps

Force free diffusion model to float32 on CPU

Add free local diffusion dependencies

Use free local Diffusers image model

Add huggingface_hub dependency

Use HF Inference Providers

Show Hugging Face model errors

Report actual model sources

Document AI model pipeline