Long Video Reasoning and Sports Highlights Generation | Multimodal Weekly 105

​​​​​​​In the 105th session of Multimodal Weekly, we feature a paper on long video reasoning and an application to generate sports highlights.

​​✅ Woongyeong Yeo and Kangsan Kim will present WorldMM, a novel multimodal memory agent that constructs and retrieves from multiple complementary memories, encompassing both textual and visual representations.
Woongyeong Yeo: https://wgcyeo.github.io/
Kangsan Kim: https://kangsankim07.github.io/
WorldMM: https://worldmm.github.io/

​✅ ​Prashant Trivedy will present GameCut, an AI-powered video highlights for sports creators to turn raw footage into viral content in seconds.
Connect with Prashant: https://www.linkedin.com/in/prashantt...
Check out GameCut: https://www.gamecut.ai/

Timestamps:
00:07 Introduction
04:45 Woongyeong starts
05:40 Understanding long video
07:10 Scaling to day-long video
08:00 Understanding videos with text
09:45 Multimodal memory
11:15 WorldMM
11:52 Episodic memory
13:40 Semantic memory
17:05 Visual memory
18:35 Adaptive memory retrieval
21:00 Kangsan discusses main results
24:20 Effect of memories
27:00 Additional analysis and efficiency analysis
29:30 WorldMM takeaway
33:25 Prashant starts
34:47 The challenges with sports highlight generation
36:10 The GameCut solution
37:40 How it works
39:20 TwelveLabs integration
40:30 Demo
54:52 Wrap Up
58:45 Q&A with Prashant

Join the Multimodal Minds community to receive an invite for future webinars:   / discord