Real-world data
for robotics AI.

Speech, first-person video, and field footage, collected fresh to your spec by real people in real places. Consented and quality-checked.

Collected for your model, not resold to everyone.

Off-the-shelf datasets are sold to many buyers at once. We collect new data to your exact spec, so what you train on is yours.

See the comparison

Off-the-shelf datasets

  • Same clips licensed to many of your competitors
  • Fixed scenes, mostly from North America and Europe
  • Collected years ago for someone else's use case
  • Unclear consent and rights history

Custom collection with Masana

  • New footage and recordings made for your project
  • You choose tasks, locations, devices, and contributors
  • Real homes, streets, and workplaces public datasets rarely cover
  • Delivered in days to weeks, starting with a small pilot
  • Signed consent from every contributor, on file

From human hands to robot hands.

Robots learn physical skills by watching people do them. Our footage is the lesson.

Human demonstration
MASANA H-1 / FRONT VIEWPICK, HANDOFF, PLACE 1HEAD CAMERA 2SHOULDER, 2 DOF BIN AHANDOFFBIN B WORKCELL Robot policy
  1. People do real tasks on camera, in real homes and workplaces.
  2. We label every step so a model knows what happened and when.
  3. Robots learn the skill and practice it in new places.

What we collect

Every clip ships labeled, documented, and consented. Pick a category to see an example.

Sample clip coming soon
    Clip metadata
    Capture
    Format
    Labels

    Example metadata shown. Delivered as MP4 or WAV with JSON metadata and a CSV manifest, or RLDS and LeRobot on request.

    Ways to work with us

    For AI teamsCustom collection

    A dataset built to your spec, starting with a small pilot.

    Request a pilot     For AI teamsOngoing data program

    Recurring batches as your model grows: new tasks, locations, and languages.

    Plan a program     For data companiesWhite-label partner

    Data you can't source in-house? We capture it under your brand and QA. You keep the client.

    Partner with us     For content ownersArchive licensing

    Own a video, audio, or text library? We license it to AI companies. You keep ownership.

    License your archive
    Swipe for more

    Start small. Scale when it works.

    1. Spec

      You tell us the task, setting, volume, and format.

    2. Recruit

      We find and brief contributors who match your profile.

    3. Capture

      Footage and recordings made in real homes, workplaces, and streets.

    4. Check and label

      Every file reviewed against your spec, then labeled.

    5. Deliver

      Organized, documented batches in your format.

    Swipe to see each step

    A pilot runs the whole loop on a small batch, usually in 1 to 3 weeks. Your feedback updates the capture guide for every batch after it.

    Questions

    Is the data exclusive to us?

    Custom collections are made for your project. Exclusivity terms are agreed in writing before we start.

    Where do you operate?

    Anywhere you need it. Our home base is Bali, and our deepest network is across Indonesia and Southeast Asia. Projects elsewhere are scoped one by one, through our own contributors or local partners. If we can't do it well, we'll tell you up front.

    How is it priced?

    Every project is quoted up front: per approved hour of video or audio, per image, or per text item. Price depends on task, location, contributor requirements, and labeling. Pilots are kept small.

    How do you handle consent?

    Every contributor signs a release in their own language, aligned with the data protection law where the data is collected (GDPR, CCPA, Indonesia's UU PDP, and others). Consent records are available to you on request.

    See all questions or read how it works.

    What does your model need to see?

    Send us the spec. We'll reply with a scope, a timeline, and a pilot proposal.

    Request a pilot