Introducing Project Stera by FPV Labs, an open data infra for embodied AI research. Project Stera includes Stera-10M, with 10M+ frames of long-horizon data with persistent state tracking, and an open-source pipeline that converts raw data into training-ready formats.

The next generation of embodied AI models needs more than pixels. It needs synchronized spatial, semantic, temporal, and action-rich knowledge in an environment turned into 4D data.

Stera is an open-source infra that turns an off-the-shelf device in your pocket into a high-fidelity multimodal data pipeline. It's built around four layers. Capture → Process → Evaluate → Export.

Stera Capture removes the need for bespoke/gated hardware and runs on an off-the-shelf iPhone. It fuses together synchronized RGB, IMU, Lidar-guided depth, and 6-DoF pose out of the box from ARKit and exports them to a raw MCAP file.

The processing layer is an open-source SDK that turns MCAP files into training-ready episodes through a unified pipeline

We are also releasing Stera-10M as an accompanying open-source dataset, comprising 10M+ frames and 200+ hours of long-duration multimodal data with persistent state tracking captured entirely using the Stera stack.

Stera-10M is a multimodal 4D dataset that includes the modalities listed below - it demonstrates what the platform can produce, is fully reproducible via the Stera stack, and stands on its own as a research-grade corpus.

The foundation for embodied AI will not be locked behind gated/bespoke hardware and will be made accessible to every researcher/builder/startup/data lab who wants to produce high-fidelity, training-ready data without having to rebuild the stack from scratch.

Read the full blog at - Stera is open today at Access Stera-10M on HF at - Arxiv Paper - FPV Labs -

10M frames is modest for embodied research at scale. Long-horizon tasks need denser state data than Stera-10M supplies

Stera is stellar.

best open-source of 2026!!

Not the best yet - bigger releases coming soon!

Just the beginning .

Small step towards the future of abundance!

lfg 🔥🔥🔥

Would love for you to try the product!

Congrats on the launch 🚀🚀

@build_plate Thank you for your contributions 🙏

Congratulations team! ✨💪

Thank you ♥️

Let's go!!

Thank you! Would love for you to try the product!

This is great 🔥

Thank you - you can setup your own fleet now :)

Congratulations 🥂 on the launch!!

Thank you♥️

Looks interesting 🔥🔥

Pls try the product and share feedback!

I'm wondering if my cat counts as embodied AI with persistent state tracking and a mixture of experts.

🚀🚀

Congrats always nailing man!

Thank you - would love for you to try the SDK - open source from India for the world :)

For sure !

hugging face hit 1M datasets in 2024. the bottleneck was never data volume.

Congrats on the release, love that ya'll opened up the SDK. Is the IOS app going to be open as well?

Thanks - ios app will follow soon :)

Amazing =] love to see it

Persistent state tracking rarely holds across 10M frames in practice. Most embodied datasets fragment well before that scale.