Looking for pilot gyms

Make grappling
footage useful.

Find positions, compare rolls, and jump to the right moment without scrubbing for hours. We are starting with BJJ and building toward better grappling analysis across MMA.

Join the pilot Read the manifesto

The platform

A timeline for every roll.

Martial Vision turns raw training footage into a simple map of the round. Search it, study it, and return to the moments that matter.

  1. 01

    Find a position

    Jump to every guard, pass, mount, or back-control sequence.

  2. 02

    See patterns

    Compare where time goes across an athlete's rolls.

  3. 03

    Make clips

    Pull out examples for class, review, or a competition plan.

  4. 04

    Keep the history

    Build a film library that becomes more useful over time.

Less time finding the footage. More time using it.

Why not just use a chatbot?

A useful timeline is different from a good-sounding answer.

A general AI tool may describe a clip well. But it was built to handle many subjects. It was not built to mark BJJ positions the same way, roll after roll.

Good for a first pass

General video AI

  • Tries to cover almost any kind of video.
  • Can give a useful first guess.
  • May change its wording or miss the exact moment a position changes.

Built for one job

A BJJ video model

  • Learns from positions labeled and checked by coaches.
  • Uses the same small set of position names every time.
  • Can be measured on athletes and rolls it has never seen.

General AI gives us a starting point.
Coach-labeled footage makes it specific.

Where the technology stands

Major U.S. sports can track almost everything.

Basketball and baseball use custom camera systems built around their sports. Combat sports are moving into AI too, but useful analysis of everyday gym footage is still early.

NBA

Body movement, tracked live

The NBA's camera system follows 29 points on each player, 60 times per second.

MLB

A system built for baseball

Every ballpark uses a 12-camera setup to track pitches, hits, players, and bats.

UFC

AI built around fight data

UFC and IBM are using fight data and language models to create live insights for fans.

The gap

Those systems are not built for an ordinary grappling room.

A martial arts gym has phone footage, changing camera angles, and athletes tangled together on the mat. It needs a model that understands grappling positions—not a stadium full of hardware.

Sources: NBA tracking, MLB Statcast, and UFC–IBM partnership.

What this could unlock

Better film review now. Better tools later.

01

For athletes

Find repeated problems and see how positions change across weeks of training.

02

For coaches

Build classes and competition plans from real moments instead of memory alone.

03

For gyms

Turn a pile of recordings into an organized, searchable team library.

04

For MMA

Start with BJJ. If it works, extend the same approach to wrestling exchanges and grappling phases in MMA.

How the model improves

The platform learns from checked examples as people use it. The work stays inside the normal review flow: fix a wrong position, save the timeline, and move on.

Privacy

Your footage. Your call.

Each athlete chooses how their footage can be used. A gym owner cannot choose for them.

Level 1

Private

Only your gym can use it.

Level 2

Research

It can help improve the system, but it is not released publicly.

Level 3

Public

Approved clips can be shared, with or without your name.

Pilot gyms

Help us test it.

For coaches

Get faster access to the moments you want to teach, study, and compare.

For gyms

Turn everyday training footage into a useful library and help shape where the platform goes next.

Contact

Interested in the pilot?

hello@martialvision.example

Replace with your real address before publishing.