List your company
ResearchFormerly WizWand

Long-form Multimodal Storytelling

0.0674BERTScoreOurs
0.0068720.0225860.03830.054014May 22, 2025
Updated 5mo ago

Evaluation results

MethodLinks
OursDescription=Full integrated design with Recursive Narrative Bank and speaker conditioning2025.050.06741.872627.482216.2412
BLIP2025.050.06741.872627.482217.0377
Speech w/o renderingRendering=False, Note=Ablation of speaker conditioning2025.050.06741.872627.482249.6302
Bark2025.050.06741.872627.482216.4668
Tacotron 22025.050.06741.872627.482216.4484
Without KeyFrameKeyFrame=False, Note=Ablation of visual grounding2025.050.03920.860925.306122.4406
VloggerNote=Integrated into dialogue and speech framework2025.050.00940.264826.170318.6999
Without RNBRNB=False, Note=Ablation of Recursive Narrative Bank2025.050.00940.526827.225916.5604
MochiNote=Integrated into dialogue and speech framework2025.050.00920.127226.610717.7431