Collected sources and patterns will appear here. Add from search, explore, or the patterns library.
List<ModalityInput> -> UnifiedConditioningVector
Project heterogeneous reference inputs (text, image, audio, video) into a unified cross-attention conditioning vector.
Problem it solves
Mixing multiple conditioning modalities directly leads to interference and weak control fidelity.
Consumes
Emits
The real projects this mechanism was found in. Attribution is the point — this is how the best teams actually do it.