Personalized Image Generation from an Author Writing Style
This addresses the challenge of visual authorial style personalization for creative assistance applications, though it appears incremental in combining existing LLM and diffusion models.
The paper tackled the problem of translating author writing styles into visual representations by developing a pipeline that uses structured author summaries to generate text-to-image prompts, achieving a mean style match score of 4.08/5 from human evaluators.
Translating nuanced, textually-defined authorial writing styles into compelling visual representations presents a novel challenge in generative AI. This paper introduces a pipeline that leverages Author Writing Sheets (AWS) - structured summaries of an author's literary characteristics - as input to a Large Language Model (LLM, Claude 3.7 Sonnet). The LLM interprets the AWS to generate three distinct, descriptive text-to-image prompts, which are then rendered by a diffusion model (Stable Diffusion 3.5 Medium). We evaluated our approach using 49 author styles from Reddit data, with human evaluators assessing the stylistic match and visual distinctiveness of the generated images. Results indicate a good perceived alignment between the generated visuals and the textual authorial profiles (mean style match: $4.08/5$), with images rated as moderately distinctive. Qualitative analysis further highlighted the pipeline's ability to capture mood and atmosphere, while also identifying challenges in representing highly abstract narrative elements. This work contributes a novel end-to-end methodology for visual authorial style personalization and provides an initial empirical validation, opening avenues for applications in creative assistance and cross-modal understanding.