Camera panel¶

The camera panel shows video frames for the currently-selected camera clip, with frame-accurate navigation, zoom-and-pan, and a right-click pixel-intensity probe.

Anatomy¶

┌─────────────────────────────────────┐
│                                     │  ← frame area (QLabel, dark #222 bg)
│          [ video frame ]            │
│              ● (probe dot if any)   │
│                                     │
├─────────────────────────────────────┤
│ Frame: 2987 / 107999 |              │
│ Time: 12.467s / 450.134s | Zoom: 3× │  ← counter label
└─────────────────────────────────────┘

The frame area has a minimum size of 320 × 240. The background is #222 (dark grey) until a frame is decoded.

Counter label¶

Below the frame, one of:

State	Text
No video loaded	`No video loaded`
Zoom == 1.0×	`Frame: N / max \\| Time: X.XXXs / max.s`
Zoom > 1.0×	`Frame: N / max \\| Time: X.XXXs / max.s \\| Zoom: N.Nx`

max = total_frames − 1 (the index of the last frame). Time is computed as frame / capture_fps.

Frame decoding¶

Video frames are decoded by a background FrameWorker (a QObject moved onto a dedicated QThread). It keeps an LRU cache of the last 32 decoded frames so small back-and-forth scrubbing is fast. Opening a new clip increments an internal generation counter; in-flight decode requests queued before the latest open_video are silently discarded.

While a frame decode is in flight, the counter updates immediately but the pixmap only refreshes when the worker replies. On clips where the cv2 capture cannot be opened, the frame area turns dark red (#7a1f1f) and shows Video unavailable: {msg}.

Zoom¶

Input	Action
Wheel up	Zoom in by factor `1.25`, centered on cursor
Wheel down	Zoom out by factor `0.8`, centered on cursor
Double-click (left)	Reset zoom to `1.0×` and pan to `(0, 0)`

Zoom range is [1.0×, 20.0×]. 1.0× means fit-to-label. Cursor-centered zoom keeps the source pixel under the cursor fixed as the zoom level changes.

At zoom == 1.0× the pan is forced to (0, 0) — you cannot pan the image off-center when it's fit to the label.

Pan¶

When zoom > 1.0×, left-mouse press-and-drag pans the image:

Drag right → image moves right relative to the label (you see more of the left side of the frame).
Pan is clamped so you can never drag the image past its own edge.
Pan is reset on double-click and on clip swap.

Intensity probe (right-click)¶

Right-click (contextMenuEvent) drops a single intensity probe dot at the clicked source pixel:

A 5-radius red circle (rgba(255, 60, 60, 220)) with a 1 px white outline is drawn on top of the frame.
dot_dropped(src_x, src_y, current_frame) is emitted; Level 2 forwards this to the IntensityWorker on its own QThread to sample ±120 frames of luma (see §5.7 Intensity plot).
Dropping the dot outside the image (e.g. in the letterbox) is silently ignored.
Dropping a new dot replaces the old one.

The dot persists through scrubbing, zooming, and panning until the clip is swapped, the user calls clear_dot() (wired to Back), or a new right-click elsewhere replaces it.

Out-of-range display¶

In Locked mode, when the MIDI drives the camera and the target frame falls outside the clip, the frame area switches to a neutral grey background (#555) with centered white text: "Camera clip starts in X.XX s" or "Camera clip ended X.XX s ago". Navigating the camera directly (e.g. wheel, click-drag pan, Left/Right) restores the normal display.

Keyboard stepping (driven from Level 2)¶

When the camera panel is the active panel:

Left / Right — step 1 frame.
Shift+Left / Shift+Right — step 10 frames.

All seeks clamp to [0, total_frames − 1].

Signals emitted¶

Signal	When
`position_changed(int)`	Any time `set_frame` or `step` updates the current frame
`user_interacted()`	On wheel, left-click, or double-click (not on programmatic updates from Level 2)
`dot_dropped(int, int, int)`	Right-click drops an intensity probe (src_x, src_y, center_frame)
`dot_cleared()`	Clip swap, `clear_dot()`, or `close_video()`