Curated page transitions from Omni-StoryBench, each showing a different aspect of the benchmark, with outputs from baseline systems. ?
Each example is a page transition. Given the current page, book-level metadata, and a condition for the next page (Input), a model generates the next page's illustration, narration, and a spoken character utterance. The book's actual next page is the Ground truth.
All 900 samples are on Hugging Face.
Loading…