AIIC AI Intelligence Centre

SOURCE-LINKED INTELLIGENCE

TokenStream: Multi-Subject Personalized Video Generation with Token-wise LoRA and Guided Attention

arXiv · AI, language, vision and robotics · article · Nov 12, 2026 · UTC

We introduce TokenStream, an optimization-based framework that composes independently learned subject identities in text-to-video generation without a reference-video training dataset or joint optimization of each new subject combination. Learning from a single reference image makes subject appearance difficult to separate from static-image bias, while merging independently trained adapters can mix identities. Token-wise LoRA (ToRA) restricts low-rank updates to the cross-attention value projections of subject-associated text tokens, providing a separate adaptation path for each subject. A T2V

Read original source ↗ Open in workspace

recordType
paper
region
Global

Evidence & attribution

First collected: 2026-11-13T16:53:25.588Z. This is not the publication date.