DissolveStereo framework generates stereo videos with improved depth and temporal consistency

Researchers have developed DissolveStereo, a new framework for generating stereo videos without requiring paired training data. The method utilizes video diffusion models and introduces a noisy restart strategy and iterative refinement to ensure temporal and spatial coherence between left and right views. DissolveStereo also employs dissolved depth maps to simplify latent space operations, leading to improved depth consistency and temporal smoothness compared to existing approaches. AI

Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →

IMPACT Introduces a novel approach to stereo video generation, potentially improving the quality and consistency of synthesized content.

RANK_REASON The submission is an academic paper detailing a new method for stereo video generation. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CV →

• DissolveStereo
• arXiv
• Jian Shi

• paper
• other