# DVIDIA — Human skill, open to everyone

> Public page text · DVIDIA

Canonical page: [DVIDIA — Human skill, open to everyone](https://dvidia.org/)

Physical AI · Built together

# Human skill. Open intelligence.

What you know today can teach the machines of tomorrow.

[Read our manifesto](https://dvidia.org/manifesto)       It starts with you

Credit the people behind the work     Keep skills open and portable     Start with the phone you have

01 / Choose your starting point

## What would you like to build?

[I know how to do something observe record **share** 01 Teach a physical skill One task. Your point of view. Make your first recording](https://dvidia.org/record)   [I want to build with demonstrations 02 Explore open skills Watch demonstrations. Find your next contribution. Explore the library](https://dvidia.org/store)        **Your first camera is already in your pocket.**

Phone or wearable. Choose your setup.

[Find your setup](https://dvidia.org/hardware)

03 / Start with what you know

## Small actions. Useful intelligence.

[Open the recorder](https://dvidia.org/record)
- 01 Record a real task Show the setup, action and result.

- 02 Review before you share Your original stays yours. You choose what to share.

- 03 Build toward a reusable skill Demonstrations → datasets → evaluated robot skills. Validation & publishing in development

04 / Your work. Your choice.

## Share the skill. Share in the value.

Your footage stays private until you choose otherwise. Our planned licensing path connects training-data buyers with creators.

[Explore workspaces](https://dvidia.org/workspace)

**Optional open releases** Choose an open license when you want others to freely reuse your work.

**Training-data licensing** Planned: buyers license selected data through DVIDIA, under agreed terms.

**Contributor revenue sharing** Planned sharing of collected license revenue with contributing rights holders. Split and terms to be agreed.

Commercial licensing and payouts are not live. Uploading alone does not grant buyers training rights.

Our vision · Shared abundance

## More time to be human.

What would you do with more of your time? Grow something. Make music. Be with someone you love.

We’re bringing human know-how together to explore what machines can learn—and make more room for the things we choose to do ourselves.

[Explore our vision](https://dvidia.org/manifesto)

A future worth contributing to

## You know something the future could use.

[Bring your first demonstration](https://dvidia.org/record)

02 / The spectrum of skill

## Every skill has a place.

From useful routines to a maker’s signature.

[How the classes work](https://dvidia.org/skill-classes)

Every class has value. Open knowledge stays open.

Open research · Public alpha

## DVIDIA Training

Train locally. Inspect what worked.

Synthetic examples, small visual predictors and simulation movement candidates.

[Watch recorded demo](https://huggingface.co/spaces/Dvidia/dvidia-training)   [Read the research](https://research.dvidia.org/papers/skillspace-footage-training/)

Saved one-scene replay. Local training runs separately. Simulation only. No human-video action recovery or hardware qualification.

[Training code](https://github.com/Dvidia-Inference/dvidia-training)

Run locally on CPU.

[Example dataset](https://huggingface.co/datasets/Dvidia/dvidia-training-examples)   [Pilot models](https://huggingface.co/Dvidia/dvidia-training-pilot)

Synthetic visual and native simulation data.

[12-layout pilot](https://github.com/Dvidia-Inference/dvidia-training/blob/main/benchmarks/sample-count-19047/README.md)

Successes, failures and matched controls.

[Follow on Hugging Face](https://huggingface.co/Dvidia)

---

[Agent index](https://dvidia.org/llms.txt) · [Markdown mirror](https://dvidia.org/mirror.md)