Dream Labs, founded by former NVIDIA robotics researchers, has announced a $90 million seed round led by B Capital and Kindred Ventures, alongside the public launch of the API for its "World Action Model" Dream-1.
Key points:
- Cross-hardware by design: Dream-1 learns from videos of people performing everyday tasks instead of being tied to a specific robot. It predicts what happens next in a scene and translates that into actions, so it can drive a two-armed robot it has never seen before with zero-shot performance — no months of tuning for a new machine or task;
- Plug and play: feed the API a robot camera image plus one plain-English instruction and it returns motor commands. The company also launched a browser-based Robot Playground where anyone can remotely drive a real robot in its San Francisco office in real time;
- Pay by the hour: Dream-1 costs $3 per hour, with a lightweight version at 75 cents; a fine-tuning service returns a tailored model within hours from a small batch of demonstration videos. The company already works with more than 20 hardware partners, and the funding will go mainly toward expanding compute capacity.
This marks embodied AI moving from demo videos to a sellable API service. Whether it can stay stable and reliable on customers' own hardware remains the real test it must pass.
Source: Read the original article. This is a summarized digest for reference only.