In this video I go from a broad game idea to a finished playable Godot prototype using a realistic, end-to-end AI workflow. I start by defining a simple high concept (a fast-paced arcade fighter where a lone street fighter defends the city from humanoid pigs), then I generate concept images with ChatGPT. The big takeaway here is iteration: the first images were way too detailed (almost like 3D renders), so I kept prompting until I got clean, simplified 2D-friendly designs for the player, the enemy, and even a matching background. After that, I animate those static images with God Mode’s AI Sprite Generator by uploading a reference image, choosing a view type (side-scrolling in my case), and generating animations one at a time (walk, punch, enemy run). I download the sprite sheets, remove the background in Photoshop (any editor works), and bring everything into Godot using AnimatedSprite2D + SpriteFrames. I also show how I use ChatGPT to help scaffold the player logic (walk/idle/punch states), plus a practical fix for mismatched sprite sizes by adjusting scale during the punch state. Finally, I break down the core game logic: a main scene that spawns enemies on a timer, enemies that move toward the player and flip direction, and a punch hit check that deletes enemies inside a radius.