Generative tapestry synthesis for long-form visual storytelling
Abstract
Recent advances in diffusion-based generative models have enabled high-quality image synthesis from textual descriptions but most existing text-to-image systems operate within fixed image boundaries or under static textual conditioning, limiting their ability to generate long-form visual artifacts that maintain both spatial continuity and semantic progression. This project investigates the problem of generating tapestry-like panoramic images that evolve narratively across space while remaining visually coherent as a single continuous scene. Initial experiments using sequential outpainting and evolving prompts revealed recurring failures such as seam artifacts, panelisation, prompt drift, and weak narrative continuity. To address these limitations, the project developed a shared-latent MultiDiffusion-based generation pipeline that produces the panorama jointly rather than extending it tile by tile. The final system incorporates multiscale coarse-to-fine refinement, structured narrative planning, and localized semantic reinforcement to guide different regions of the canvas towards distinct story beats while preserving global continuity. A prototype implementation was developed and evaluated qualitatively through iterative refinement and comparative experiments. Results show that the proposed architecture substantially improves seam quality and wide-scene coherence as compared to earlier sequential outpainting baselines while demonstrating potential for spatially varying…
Read the paper
Related studies
- Learning from First-Generation Qualitative Approaches in the IS Discipline: An Evolutionary View and Some Implications for Authors and Evaluators (PART 1/2)
- Commentary on the clinical and preclinical dosage limits of interstitially administered magnetic fluids for therapeutic hyperthermia based on current practice and efficacy models
- Prosthesis: New Technological Opportunities and Innovative Biomedical Devices
- The history and future of neural modeling for cochlear implants
- Ryanodine receptors: waking up from refractoriness
- Novel Transgenic Rabbit Model Sheds Light on the Puzzling Role of Matrix Metalloproteinase-12 in Atherosclerosis
Go deeper in Evidence Hub
Evidence Hub adds a study data sheet (design, sample size, evidence level, key numbers), an AI summary and a chat grounded in the paper's text.