video-kenburns — Traditional AI Model

Details

• CategoryTraditional
• Year-1
• License

Inputs & Outputs

ImageInput · image

Still image to animate

VideoOutput · video

The still, moving

Parameters

durationfloat · default 5 · range 1–20

Shot length in seconds.

fpsinteger · default 30 · range 12–60

Output frame rate.

Output frame rate. A slow move has more room to judder at a low rate than a high one, so 30 is the default rather than 24; there is no source rate to match here, since the input is a still.

movestring · default Push In · options: Push In, Pull Out, Pan Left to Right, Pan Right to Left, +3 more

Camera move. 'Push In' and 'Pull Out' zoom on the centre; the pans hold a fixed window size and travel across the image. Ignored when start_rect / end_rect are set.

Named camera move. 'Hold' delivers a static frame (still useful — it renders the still to a video the rest of the chain can concat). 'Push In'/'Pull Out' ease a centred zoom over the whole shot. The four pans hold the window size and travel it across one axis. 'amount' controls how far every preset travels. All presets are shorthand for a (start_rect, end_rect) pair — set those directly when the shot needs a window no preset can name.

amountfloat · default 0.12 · range 0.02–0.6 · advanced

How far the move travels, as a fraction of the frame. Higher is more visible — and on a zoom, the tight end is enlarged by this much.

Travel distance as a fraction of the frame. The default (0.12) is deliberately larger than Video-Reframe's whole-clip 0.08: a live clip has its own motion to keep the shot alive and a still has none, so this drift is the only thing between the shot and a freeze-frame. Note the cost on zooms — the tight end of a Push In is enlarged by this fraction, so a big value on a small scan softens the close end. Ignored when start_rect / end_rect are set.

start_rectstring · default · advanced

Opening window as 'x,y,w,h' in fractions of the source (0-1). Leave empty to use the move preset.

Authored opening window, written as 'x,y,w,h' in fractions of the SOURCE image — fractions rather than pixels so the same shot list survives a re-scan at a different resolution. Setting either rect overrides 'move' and 'amount' entirely. The rect is GROWN to the output aspect ratio, never cropped, so the delivered framing is a superset of what you asked for: you can name the region that matters without also knowing the canvas shape. Set start_rect alone to hold on one detail for the whole shot.

end_rectstring · default · advanced

Closing window as 'x,y,w,h' in fractions of the source (0-1). Falls back to start_rect when only one is given.

Authored closing window in the same 'x,y,w,h' source fractions as start_rect. Give both to travel from one framing to another (open wide on the whole painting, close on a face). Give only one and the shot holds on it — the missing end falls back to the one you named rather than to a preset, so an authored window never travels somewhere nobody asked for.

fitstring · default Fill · options: Fill, Fit (Matted)

'Fill' crops the window to the canvas with no bars. 'Fit (Matted)' shows the whole window on a flat mat — use it when the complete image has to stay visible.

How the window meets the canvas. 'Fill' crops to the canvas edge-to-edge and builds no background at all, so no mat colour can reach the frame. 'Fit (Matted)' scales the whole window to sit inside the canvas and pads with a flat colour — the only way to show a complete landscape painting on a 9:16 canvas without cutting it. In Fit the window keeps the SOURCE aspect for the whole move, so the mat stays one size instead of breathing.

backgroundstring · default Hanji · options: Hanji, Ink, Black, White

Mat colour behind a matted shot. Only applies to 'Fit (Matted)'.

Flat colour filling the bars in Fit mode. 'Hanji' is rgb(191,177,144), measured as the median paper tone across the 30 Hyewon scans — a warm mat reads as a mounted work where a cool one reads as a screenshot on a slide. 'Ink' is a near-black gallery wall, 'Black' is true black for OLED delivery, 'White' for print-style layouts. There is deliberately no blurred-copy option: a blurred duplicate of the artwork is still a derivative of it, which is exactly what this node exists to avoid.

output_sizestring · default 1080x1920 (9:16) · options: 1080x1920 (9:16), 1080x1350 (4:5), 1080x1080 (1:1), +1 more

Output canvas.

Delivery canvas. 9:16 is the Shorts/Reels default; 4:5 and 1:1 are the feed crops; 16:9 renders a landscape master from the same authored rects without re-planning the shot list.

easingstring · default Ease In-Out · options: Ease In-Out, Linear · advanced

Motion curve. 'Ease In-Out' starts and stops gently; 'Linear' moves at a constant rate.

Motion curve across the shot. 'Ease In-Out' is a smoothstep — the same curve Video-Reframe's camera moves use, so a flow that intercuts stills with live clips does not have the two moving at visibly different rates. 'Linear' starts and stops at full speed, which reads as mechanical and is mostly useful for a continuous pan meant to be cut mid-move.

Tags

• video
• image
• ken burns
• pan
• zoom
• still
• slideshow
• archival

Alternatives in Video Editing

• Audio Finish

  Normalize a video's loudness to a broadcast target (2-pass loudnorm + peak limiter).

• Hook Teaser (Cold Open)

  Prepend a cold-open teaser of each clip's hook moment to the front.

• Narration Mux (Cold Open)

  Prepend a voiceover cold-open (frozen first frame + narration) to a clip.

• Remove Dead Zone

  Remove silent gaps from each clip in a video list to tighten pacing.

• Slice Transcript to Clips

  Slice a whole-video transcript onto each clip.

• Snap Segments to Speech

  Snap segment boundaries to speech word boundaries.

• Video Caption

  Burn each clip's highlight text onto it as a caption overlay.

• Video Clean Span

  Trim each clip to its single dominant continuous shot (remove mid-clip frame jumps).