...
Turn a single book title (and an optional story you already wrote) into a finished, illustrated children's picture-book PDF with a consistent protagonist across every page. An LLM splits or writes the story page by page and invents a Master Style Guide plus a protagonist description; a text-to-image model renders one "anchor" portrait of that protagonist; then every page is generated image-to-image against the anchor so the character's face, hair, clothing, and signature accessory stay locked from cover to last spread; finally the cover + per-page text/image spreads are bound into a downloadable PDF. Two entry modes: give only a title and the model writes the whole story in a warm read-aloud bedtime voice, or paste your own story (with optional `## Page N` markers) and it preserves your wording, only tidying and splitting into pages. Every page also seeds "find me" hidden-detail motifs for rereadability. Use this when someone wants to make a kids' picture book / bedtime storybook end to end and walk away with a print-ready PDF, not just loose text or stray images. Input is a title plus page count (and optionally your own story text and a choice of illustration model); output is a bound illustrated picture-book PDF with a visually consistent main character.
创建者: songlin she · GitHub