Results
Ranked from 11 ratings. Score is adjusted from raw score by the median number of ratings per game in the jam.
The Biggest Challenge
Challenge: AI makes things quickly, but it does not know what the game is about. The generated art, the synthetic voice and the code each failed in ways I had to catch myself: keyed images with half-dissolved edges, a voice effect that made every clip sound the same, and a pumpkin animation that ran one beat behind the text. Every fix was my decision.
Breakthrough: deciding that nothing generative runs while you play. The game is a deterministic state machine with no runtime LLM and no API. The AI built the tools and assets; I wrote every line of dialogue and designed every rule. That kept the voice of the game mine and made it testable. I also overruled an AI suggestion: it wanted the voice synced to the planchette, and I chose to keep four channels deliberately out of step, because the gap is where the unease comes from.
AI Tools Used
Nano Banana (board and pumpkin art, cover art), CapCut TTS (the Father's voice, 118 clips recorded and named by me), Claude (code, build scripts, and machine translation drafts into Japanese, Chinese, Korean, German, Spanish and French).
Areas Where AI Was Used
Images: board, seven pumpkin faces, cover. “cover art (Nano Banana, title added by me and Claude)”. Sound: the Father's voice (synthesised speech); ambient sound is live Web Audio code. Code: engine, board, input, build and translation scripts. Translation: six languages, machine-made and labelled as such. Writing: the dialogue, the roles, the techniques and the design are mine. A few lines (some refusal lines, one break line, some stage notes) began as Claude drafts, and I'm rewriting them in my own voice. Nothing generative runs at play time. The full who-made-what table is in my Dev Log.
Leave a comment
Log in with itch.io to leave a comment.