Skip to content

MIDI–Camera Alignment Tool

A PyQt5 desktop tool for temporally aligning overhead camera recordings (Sony FX30, ~240 fps) with Disklavier MIDI files across a multi-participant piano study.

The two recording systems run on unsynchronized clocks separated by a constant 1–20 minute offset per participant. This tool provides a manual two-phase workflow to recover the alignment and persist it as JSON.

Who this is for

Researchers working with paired Disklavier MIDI + overhead video recordings who need to index MIDI events against camera frames for downstream analysis.

What it does

  1. Scans a participant folder for .mid files in disklavier/ and .MP4 + .XML pairs in overhead camera/.

  2. Displays every MIDI clip and every camera clip as a Gantt-style timeline (Level 1).

    Level 1 timeline overview showing MIDI and camera bars on a shared time axis

  3. For any MIDI + camera pair, opens a detail view with a falling-keys piano roll, a frame-accurate video viewer, a dual-track overlap bar, a pixel-intensity probe, and an anchor table (Level 2).

    Level 2 detail view showing the MIDI piano roll, camera panel, overlap indicator, intensity plot, and anchor table

  4. Lets you mark one MIDI keystroke and the matching video frame, then either computes a global shift or records a per-clip anchor to refine alignment.

  5. Saves and loads the session as a versioned JSON file (Schema v1).

No audio, no export

The tool does not play audio and does not export aligned data. The JSON save file is the output — downstream code consumes the anchor set and the global shift to map MIDI seconds ↔ camera frames.

Where to start