Dripta Raychaudhuri

Member of Technical Staff @Physion_Labs | prev @AmazonScience

[Dripta Raychaudhuri @dripta_rayc]

the quality vs cost difference here is pretty interesting, especially with @runwayml and @openart_ai ending up so close on overall quality

really excited to see Physion-Arc Ads & Marketing out in the wild 🚀

[Physion Labs Official @Physion_Labs]

🚀 𝐏𝐡𝐲𝐬𝐢𝐨𝐧-𝐀𝐫𝐜 1.0 𝐀𝐝𝐬 & 𝐌𝐚𝐫𝐤𝐞𝐭𝐢𝐧𝐠 𝐓𝐫𝐚𝐜𝐤 𝐢𝐬 𝐥𝐢𝐯𝐞.

For this track, we wanted to test something very practical: can today’s video agents actually take a real advertising brief and turn it into a strong finished ad?

We worked with advertising and creative professionals to build 100 real-world briefs across text-only, single-reference, and multi-reference tasks, then evaluated each agent across 24 metrics covering brief fidelity, production quality, and ad effectiveness.

The first leaderboard includes @runwayml, @openart_ai, @Creatify_AI, @LumaLabsAI, and @HeyGen.

One result we found especially interesting: Runway and OpenArt AI are very close on overall quality — 71.1 vs. 69.5, with no statistically significant difference in this round.

💰 At the same time, the agents show quite different cost profiles. OpenArt AI is about $6/video, compared with $37/video for Runway, which makes the quality–cost tradeoff another interesting dimension to look at alongside overall performance.

Full leaderboard here: https://t.co/S13zr74evo

[Physion Labs Official @Physion_Labs]

🎬 Video agents can now generate minute-long videos. But can they actually direct?

Today, we’re launching 𝐏𝐡𝐲𝐬𝐢𝐨𝐧-𝐀𝐫𝐜 1.0, a new benchmark evaluating complete, multi-scene videos across narrative coherence, cinematic language, and production quality.

We tested @runwayml, @LumaLabsAI, @MiniMax_AI, @Kling_ai , @UtopaiStudios and @TapNow_AI on 100 screenplays and 600 generated videos.

🏆 𝐑𝐮𝐧𝐰𝐚𝐲 𝐀𝐠𝐞𝐧𝐭 2.0 𝐫𝐚𝐧𝐤𝐞𝐝 𝐍𝐨. 1 𝐨𝐯𝐞𝐫𝐚𝐥𝐥 𝐚𝐧𝐝 𝐥𝐞𝐝 𝐞𝐯𝐞𝐫𝐲 𝐞𝐯𝐚𝐥𝐮𝐚𝐭𝐢𝐨𝐧 𝐝𝐢𝐦𝐞𝐧𝐬𝐢𝐨𝐧. Its advantage was especially clear across subjective metrics, where cinematic taste matters most. Runway ranked first on all eight.

🔗 Read the full benchmark: https://t.co/PuBpmwUQGv

[Physion Labs Official @Physion_Labs]

We feel your pain with video generation glitches: face drift, melting hands, outfit changes, and physics that looks fine until the second watch.

AI video is getting magical — but production doesn’t happen at first glance. It happens through rewatches, edits, reviews, clients, directors, and audiences, where every frame has to hold up.

That’s why we’re opening the interactive demo for 𝐏𝐡𝐲𝐬𝐢𝐨𝐧-𝐀𝐭𝐥𝐚𝐬 1.0: to make video generation failures visible, inspectable, and grounded in real evidence.

You can interact with the demo yourself: inspect videos, reveal hidden glitches, and see how many failures you can successfully spot.
- Interactive demo: https://t.co/LPT6U5MQ6J
- Blog: https://t.co/1Ck3CpuJk6

We show an apples-to-apples view across 𝐒𝐞𝐞𝐝𝐚𝐧𝐜𝐞 2.0, 𝐕𝐞𝐨 3.1, 𝐊𝐥𝐢𝐧𝐠 3.0, 𝐇𝐚𝐩𝐩𝐲𝐡𝐨𝐫𝐬𝐞 1.0, 𝐇𝐚𝐩𝐩𝐲𝐁𝐚𝐧𝐚𝐧𝐚, 𝐏𝐢𝐱𝐯𝐞𝐫𝐬𝐞 𝐕6, and 𝐆𝐫𝐨𝐤 𝐈𝐦𝐚𝐠𝐢𝐧𝐞 — surfacing the subtle, production-critical issues that are easy to miss but hard to ignore.

Come see where generated worlds start to break 👀

[Dripta Raychaudhuri @dripta_rayc]

@saptarshipal_ @MPI_EvolBio @chilbe3 Congrats!

[Rohit Lal @take2rohit]

🛑New Paper Alert!🛑
Introducing POISE for human silhouette extraction! 🤖✨ Our latest research tackles the tricky problem of occlusions with a novel self-supervised approach. (1/5)

[Dripta Raychaudhuri @dripta_rayc]

@KostasKarydis @ucr_ece @UCRBCOE @UCRiverside Congratulations!

[Dripta Raychaudhuri @dripta_rayc]

Come check out our work on Controllable Dynamic Multi-task Architectures at #CVPR2022 oral session 3.1.2 (8.30-10.30am) and poster session 3.1 (10.00-12.30pm) tomorrow June 23!

Project page: https://t.co/qkDp5nAaRn

[Michael Black @Michael_J_Black]

I get a lot of reviews that say my work is not novel and I bet I'm not alone. It's always frustrating because I see novelty where the reviewer doesn't. Rather than rebut every critique, I've written a blog post to help reviewers think about novelty. https://t.co/UXLabOkYcn

[Andrej Karpathy @karpathy]

Errr ok wow, I am shook by the new ConvMixer architecture
https://t.co/crUMktQ0ig "the first model that achieves the elusive dual goals of 80%+ ImageNet top-1 accuracy while also fitting into a tweet" 😐