Six ways to put motion in a unit, all showing the same teaching moment.

A real teacher, real body, real voice — and it costs nothing. Embedding is what YouTube's terms are designed for, so there's no licensing problem as long as we embed rather than download and re-host.

Framed as a teaching exercise it stops being filler: "watch six teachers cue this same pose, and notice they don't agree about the knee." That's a competency our standard actually asks for — reading a cue critically instead of repeating it.

What it can't be is the spine of the course. It's someone else's curriculum inside ours: their cues, their branding, their ads, and their right to delete the video on a Tuesday.

• Cost$0
• LicensingFine if embedded. Not fine downloaded or re-hosted.
• ControlNone. Can disappear, get monetised, or contradict our standard.
• ShownSix public tutorials, embedded via YouTube's player

The same figure the player already rotates, rendered out as continuous video: a full orbit of the pose in under three seconds. Built from the reconstruction of a real photograph, so the geometry is a real body's, not an artist's guess.

It's ours forever, costs nothing per unit, and can show any angle — including the ones no camera was on. It also looks like what it is: a model, not a person.

• Cost$0 per pose after the rig exists (~1 hour of compute per pose)
• LicensingOurs outright
• ControlTotal — any angle, any speed, any annotation

One pose, five heights, 5'0" through 6'4", with the consequence written on the frame: the cue "step about a leg's length apart" comes out as 107 cm on one body and 136 cm on another. The amber marker is the single fixed number a filmed-once course teaches — watch it be wrong for four of the five people.

Filming this would mean five hired bodies, five setups, one day — and re-shooting the whole thing for the next pose.

This is the clip I'd put on social media. It makes the argument in six seconds without a word of copy: your training was filmed on one body, and it wasn't yours.

• Cost$0 per pose — already generated for every proportion combination
• LicensingOurs outright
• ControlAny height, any build, any limb ratio

A real body, and the teacher's eye drawn onto it in the order a teacher actually uses it: find the ankle line, find the knee, see the drift, then check the shoulders. Seven seconds, no narration needed.

This is the format that most looks like teaching rather than content. It's also the cheapest thing on this page to produce at scale — the marks are generated from the pose data we already extract.

• CostMinutes of compute per clip. No filming, no licensing.
• LicensingOurs, on photographs we own or have licensed
• ControlTotal — every mark is data-driven, so it can be regenerated per body

Generative video is good at atmosphere and bad at joints. Held asymmetric poses are its worst case: limbs merge, feet rotate, a knee crosses the ankle mid-clip. For a program whose entire claim is we teach you to see joint position accurately, a clip where the anatomy is subtly wrong is worse than no clip.

There's a second problem that isn't technical. We're selling a credential built on real bodies, reviewed by real humans. Generated instructors are the thing our whole pitch is against.

• CostRoughly $0.30–$1.50 per 5-second clip, plus retries — and pose work needs many retries
• LicensingProvider terms vary; some train on unlicensed video
• ControlLow. You prompt, you hope.

That's the gap nothing else on this page fills. Teaching is diagnosis, and you cannot teach diagnosis with footage of things going right. We need three or four bodies deliberately demonstrating the errors our standard asks candidates to spot, filmed from the angle the error is visible from.

One day, one space, three or four models, and we own it forever — and every clip can then have layer D generated on top of it, which is where this gets genuinely good.

• Cost$2,000–6,000 for a day: space, 3–4 models, an operator, releases
• LicensingOurs outright, with signed releases
• ControlTotal — we direct the errors