Measurements

The Measurements panel is where you pick which body channels matter for your movement. Channels become per-frame numbers that drive phase transitions and form rules.

Two ways to drop a measurement

A. Preset chips (recommended for common moves)

The top of the Measurements panel shows a grid of curated chips:

Group	Chips
Stance	wide · narrow
Hand spread	together · apart
Hand height	above head · at chest · by side
Body in frame	on left · centred · on right
Knees	deep · straight · deep right · straight right
Elbows	bent · straight · bent right · straight right
Body posture	stood up · crouched down · hands raised · hands lowered · jumped up

Tap a chip → drops the underlying tracking point(s) + a pre-tuned condition into the currently-selected phase. The chip lights green (“applied”).

The chips are wired through MovementIntents (package:pose_flow). Each intent maps to one or more TrackingPoints + a default condition range, the trainer doesn’t have to remember “knees deep = left_knee + right_knee both between 70° and 110°”.

B. Channel catalogue (for everything else)

Below the chips, a scrollable list of every available channel, grouped by kind:

Joint angles: left_knee, right_knee, left_hip, right_hip, left_elbow, right_elbow, left_shoulder, right_shoulder, spine, …
Camera-frame angles (view-dependent, spine_lean, left_knee_valgus, hip_tilt, …). Marked with a 2D badge to warn that the value changes with camera position.
Distances: stance_width, wrist_spread, left_wrist_to_shoulder, …
Ratios: stance_to_shoulder_ratio, wrist_to_shoulder_spread_ratio, …
Position: hip_center_x, left_wrist_y, …
Velocity: derived channels with a 200ms window.
Stability: derived channels with a 1s window.
Diagnostics: symmetry_* channels (informational only, not currently conditionable).

Tap the checkbox on any row → the channel is enabled as a condition on the current phase. The row shows the live value to the right.

Long-press the ⓘ button → opens an info dialog explaining how the channel is computed + which landmarks it uses.

Live values

Every row in the catalogue shows the live value updated each frame, with a coloured RAG dot indicating compliance with the authored range (green = in, red = out, no dot = no range authored).

Lets you author “I want this number to be < 100” with confidence , you can see the number drop as you move.

Decimal precision

Different channels round differently:

Angles & symmetry, whole degrees
Distances & ratios, 2 decimals
Position, 2 decimals (it’s in [0, 1] normalised space)
Velocity, 2 decimals
Stability, 3 decimals

The Studio reads lookupChannelKind for each channel to pick the right format.

What gets written when you save

Each enabled channel becomes a TrackingPoint entry in the .pose file’s trackingPoints array, plus a FormRule (if you authored a range) and a PhaseCondition (if the channel gates a phase transition).

The Studio de-duplicates, if the same tracking point is referenced by conditions in multiple phases, it’s emitted ONCE in trackingPoints and referenced by id from each PhaseCondition / FormRule.

Channel selection tips

Prefer view-invariant channels (joint flexion angles, distances, ratios) when the movement doesn’t depend on camera position. These work the same at every bucket.
Use ratios over raw distances when you want body-size independence. stance_to_shoulder_ratio is robust across body sizes; stance_width (an absolute distance) is not.
Use position channels for “where in the frame” gates (lateral shuffles, predator jacks).
Use relativeToBaseline (Δ from phase entry toggle in the conditions panel) for jump detection or any “moved by X from the phase start” gate.
Camera-frame angles (spine lean, knee valgus, hip tilt) only make sense when you author per-bucket bands, they read differently from each camera angle. See camera buckets.