To bring generalist intelligent robots to the real world, we have to overcome the data scarcity problem.
At Rhoda, we are solving it by reformulating robot policies as video generation.
Today, we introduce the Direct Video-Action Model (DVA)
🙏 We’re incredibly grateful to everyone who joined our @RhodaAI party last night at @CVPR.
The turnout exceeded anything we expected, and it was a pleasure meeting so many researchers and builders from the vision / robotics community. Thank you for all the great conversations!
I'll be at CVPR next week (6/3–6/7). If you’re working on or exploring opportunities in video models for robotics (research or engineering), happy to chat 🤖
We’re also hosting a Rhoda party Thursday night with many of our technical team in town. DM me for an invite 🍻
Can a large foundation video model run as a real-time robot policy at the edge, on a single RTX 5090?
• ✅ No quantization
• ✅ No distillation
• ✅ Full denoising (all the way from noise to clean video)
We just proved it's possible. 👇🎬
How? Existing video models aren't optimized for real-time inference.
Instead of fine-tuning off-the-shelf video models, we co-design inference-aware model architectures and model-aware inference optimizations from the ground up.
Teaching a robot a new task typically means stopping operations, collecting teleoperated demonstrations, and retraining. That process takes hours at a minimum. We wanted to know if we could collapse it to seconds — from a single human demo, on the fly, no retraining required.
How it works: we train on paired human demo and robot execution data. Because our DVA, FutureVision, has long-context visual memory built in (x.com/RhodaAI/status…), we prepend the full human video into the model's context and predict robot actions closed-loop. The model
Here’s something we’ve never seen done before.
Real-world tasks are long and ambiguous. Solving them requires visual memory and state tracking. Most robot policies only see the last few frames. Ours doesn't.
We put our DVA, FutureVision, to the perfect testbed: the shell game
The future we're building toward is one where robots adapt to new tasks in seconds.
At Rhoda, we tackle real-world problems through fundamental research.
Full story + technical deep-dive:
Here’s something we’ve never seen done before.
Real-world tasks are long and ambiguous. Solving them requires visual memory and state tracking. Most robot policies only see the last few frames. Ours doesn't.
We put our DVA, FutureVision, to the perfect testbed: the shell game
How? Our DVA implements robot policy as future video generation.
Given the context, the model generates future videos (bottom left) predicting not just the correct cup to pick up, but even the appearance of the hidden object.
Native training on long, continuous videos gives the
"I don't think the world is going back to non video based pretraining." Our CEO @startupjag spoke with @bheater at @a3automate on why video is the foundation for robots that actually work in production.
1/ We are speed running industrial robotics.
It took us just 19 days from the first day of data collection to filming a 2.5-hour continuous run of our model autonomously breaking down industrial containers — zero human intervention.
The data efficiency of our DVA model is
3/ Achieving a 100% autonomous rate in a 2.5-hour continuous run means the model needs to handle all kinds of edge cases. Whether it's pulling a drifted box back into range or re-attempting a failed flip, the model self-corrects in real-time.
-> The trash is out of reach. The
Trained on just 11 hours of robot data, our model is surprisingly robust, thanks to web-scale pre-training.
It doesn't just avoid errors; it handles them. If the lid tears off, it finds a new way to grip. If a bearing is stuck, it shakes the bag loose.
Watch our robot navigate
