EPIG: Emotion-Based Prompting for Personalised Image Generation

EPIG (Emotion-Based Prompting for Personalized Image Generation) introduces a novel prompting method that enhances emotional expressiveness in text-to-image diffusion models without requiring retraining. By utilizing psychologically informed emotion representations (valence-arousal) and structured prompt enrichment, EPIG achieves significant reductions in mean arousal error—up to 17% on subject-sensitive prompts—compared to baseline methods. This lightweight, training-free approach is particularly beneficial for resource-constrained scenarios, enabling more emotionally coherent image generation that aligns with nuanced user intent.