51f93f9688
Three new nodes (SelvaModelLoader, SelvaFeatureExtractor, SelvaSampler) vendoring selva_core from jnwnlee/selva. Pure PyTorch, no subprocess, zero new pip dependencies. TextSynchformer provides text-conditioned sync features for improved audio-visual alignment. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>