Direct, Parallel, or Sequential? A Comparative Study of Training-Free Multi-Subject Image-to-Video Generation
2026-08-24 • Computer Vision and Pattern Recognition
Computer Vision and Pattern Recognition
AI summaryⓘ
The authors study how to generate videos from text descriptions when there are multiple subjects in the scene. They compare three methods: direct generation that creates the whole video at once, parallel generation that creates each subject separately and then combines them, and sequential generation that builds the background first and then adds subjects one by one. Their tests show how each method better keeps how subjects look, moves, and interact, but also reveal some weaknesses like ordering sensitivity or less subject interaction. The authors provide useful guidance for creating multi-subject videos without needing extra training.
image-to-video generationtext-conditioned generationmulti-subject videoappearance preservationmotion fidelitytemporal consistencyspatial interactionstraining-free methods
Authors
Yanliang Qi, Kexi Chen, Muchao Ye, Haomiao Ni
Abstract
Text-conditioned image-to-video (I2V) generation has advanced rapidly, yet generating videos with multiple subjects remains challenging. A model must simultaneously preserve the appearance of each subject, assign distinct motions, and maintain coherent spatial and temporal interactions. This paper presents a systematic study of three representative paradigms for training-free multi-subject I2V generation: direct, parallel, and sequential generation. Direct generation applies a pretrained I2V model to the complete reference image and prompt, requiring all subjects and motions to be synthesized jointly. Parallel and sequential generation instead decompose the reference image and prompt into subject-specific visual and textual conditions. Parallel generation synthesizes each subject independently and subsequently composes the resulting videos, reducing the complexity of each generation step at the cost of weaker inter-subject context. Sequential generation first synthesizes a background video and then progressively introduces individual subjects. This preserves accumulated scene context but introduces sensitivity to subject ordering and error propagation. We empirically evaluate the three paradigms across diverse multi-subject scenes, comparing appearance preservation, motion fidelity, temporal consistency, and inter-subject coherence, while also characterizing their distinct failure modes. Our findings reveal the strengths and limitations of each paradigm and offer practical insights for designing controllable multi-subject video generation systems.