📄 Check out the paper: arxiv.org/pdf/2412.05275
🎥 For more results: motionflow-diffusion.github.io

We’d love to hear your thoughts or questions.
Stay tuned for more from our team! 🚀

📊 How does it compare? MotionFlow outperforms state-of-the-art methods like DMT, VMC, and others by balancing:
✅ Motion Fidelity (realistic motion)
✅ Temporal Consistency (smooth transitions)
✅ Text Alignment (prompt accuracy)

🧠 Core innovation? Instead of extensive training or fine-tuning, we extract motion dynamics from attention maps during test time.

This ensures:
✅ Fidelity to original motion
✅ Freedom to edit scene layout

MotionFlow is a novel invert-then-generate framework that uses cross-attention maps to transfer motion dynamics across contexts. It maintains motion fidelity while enabling scene edits. No extra training needed.

🌟 Exciting news! Our research, MotionFlow, is out! 🚀

A training-free method for motion transfer using pre-trained video diffusion models. 🎥✨

Big thanks to Hidir Yesiltepe, Connor Dunlop, and my advisor Pinar Yanardag. More from our lab soon—stay tuned! 🧵👇

Thanks for the answer. I was expecting to hear a similar use case.

bsky.app/profile/tmer...

I am asking this because I hear a lot of people trying to get into the first page of this list so hard. Can anyone elaborate what is the benefit of getting there?

I am curious If there is anyone who reads papers from the arxiv recent papers lists. Like directly from this link: arxiv.org/list/cs/recent

I don't. But who does? Really

Thanks to all for following me. I think I somehow got into some of the starter packs. I am normally not too active on these sites but I will try to be 🙌

The worst part was that the paper was about "fonts". Vikas handled the situation quite well 😅

This one is more comprehensive go.bsky.app/Pxcnfu6

Found this one. bsky.app/profile/sedi...

cool pack, thank you 🙌

Is there any starter pack for Vision GenAI people?