Coherence-Oriented Dream Scene Visualisation

2026-08-05

AI Summary

This work addresses the challenge of intuitively conveying dream narratives by proposing the first four-panel coherent visualization method tailored for dream descriptions. The approach leverages large language models to segment dream texts into four temporally ordered fragments and employs text-to-image generation models to synthesize corresponding visual frames. To ensure semantic fidelity between text and image as well as cross-frame visual consistency, a multimodal feedback mechanism—integrating CLIP, DINOv2, and Qwen2-VL—is introduced, guiding an iterative regeneration strategy. Evaluation on 50 dream sequences from the DreamBank dataset demonstrates that the generated visual narratives achieve strong objective performance in terms of quality, faithfulness, and coherence.

Abstract

Dreams can be emotionally intense but difficult to communicate. We describe the Dream Scene Visualiser (DSV) system which turns written dream descriptions into a temporal sequence of four panel images visualising the dream. This starts with a large language model prompted to split a dream description into four chronological parts. Then a text-to-image model produces images for each part with visual coherence maintained across the sequence, and DSV regenerates any image not suitably matching the text. We evaluate DSV over 50 visualisations from dream descriptions in DreamBank, and report quality, fidelity and coherence results via objective measures employing the CLIP, DINOv2 and Qwen2-VL vision-language models.

Problem

• dream visualization
• visual coherence
• temporal sequence
• dream description
• image generation

Innovation

• dream visualization
• visual coherence
• text-to-image generation
• temporal sequence
• vision-language evaluation