Does it support multi images reference?

[Vickeyhw]

In some application scenarios, it is necessary to generate videos with two subjects. The current approach is to first use image editing to place both subjects in one frame, and then generate the video using the first frame as a control image. However, the ability to maintain ID consistency is weak and the results are not ideal. Will you consider supporting multi-image reference for better ID consistency?

[KBRASK]

Multi-subject generation is an important and interesting task, with stronger practical applications. Single-subject generation can be seen as a special case of multi-subject generation. However, transferring methods from single-subject generation to multi-subject scenarios is limited. Extending Stand-In to multi-subject tasks is a promising solution and one of my upcoming research directions.