Filming Protocol

Beta — research prototype

Xianji is an early-stage research instrument, not a shipping product. The trial engine and its timing are verified; much of the surrounding workflow is not built yet. Tennis is configured but not runnable. See Current Limitations before planning around it.

Footage quality sets the ceiling on everything downstream. These are requirements, not recommendations.

Camera position

Receiver viewpoint only. Place the camera where the receiving athlete’s eyes would be:

• On the centre line
• 0.5–1.0 m behind the baseline
• 1.6–1.8 m height

The opponent’s full body plus racket at full extension must stay in frame for every shot. A racket that leaves the frame during the backswing removes the cue you are trying to measure.

Caution

Side-on or corner footage trains the athlete to read cues a receiver never has. It produces numbers that look valid and are not comparable with receiver-viewpoint results. The app flags non-receiver footage but will not block it.

Settings — all locked, no auto

Setting	Value
Frame rate	120 fps (60 acceptable)
Resolution	1080p
Shutter	1/1000 – 1/2000
Focus	Locked, autofocus off
Exposure	Locked
White balance	Locked
Stabilisation	OFF — both EIS and OIS
HDR	Off

Why each one matters

Stabilisation is the one people forget. It subtly warps time and space between frames — exactly what you are measuring. Turn off both electronic and optical stabilisation.

Shutter speed matters more than it looks. At 1/60 the racket is a smear across precisely the frames where the cue lives. At 1/1000 the racket face is readable frame by frame.

Autofocus hunting mid-rally blurs the frames around contact. Lock it on the opponent before recording.

Frame rate determines your temporal resolution. At 120 fps each frame is 8.3 ms; at 60 fps it is 16.7 ms, which limits how precisely a contact frame can be marked.

Note

The app reads the exact frame rate from the file. Phones record 119.88 fps, not 120.0 — contact times derive from the real value, so never override it.

Session planning

• Aim for 80–100 shots to end up with 60 usable ones
• Cover all zones roughly evenly. Missing zones narrow the athlete’s real choice and inflate accuracy
• Keep the camera fixed across a footage set. Moving it mid-set makes the shots non-comparable
• One opponent per footage set where possible — a threshold measured against one opponent’s technique is not directly comparable to another’s

Consent

You are recording an identifiable person’s technique for repeated analysis. Record their consent — the manifest has a field for it, and the app flags footage that lacks one.

Next steps

• Preparing Footage →
• Manifest Format →