talking_head_framing v1.0.0

name: "talking_head_framing" description: "Directs and edits talking-head video: shot framing, eye-line, background design with depth separation, and jump-cut rhythm that holds viewer retention. Triggers when asked to shoot, frame, direct, or edit talking-head or interview footage, or review a talking-head video for quality."

Talking Head Framing

Purpose

Turn raw talking-head footage (or a blank setup) into a shot and edit plan that looks intentional on camera and holds retention in the timeline. Covers the full chain: where the camera goes, where the eyes sit, how the background reads, and how the cuts move.

Workflow

1. Frame the shot

• Eye level, not above or below. Put the lens at the speaker's eye height. High angles shrink the speaker; low angles distort.
• Eyes on the upper third. The speaker's eye line sits on the top third line of the frame, never in the center. Leave a small headroom gap above the head: too little feels cramped, too much makes the speaker sink.
• Direct address goes centered; interviews go off-center. Looking straight down the lens? Center the speaker. Looking at an interviewer off camera? Put the speaker on one vertical third line and leave lead room (nose room) in the direction they face. Two speakers sit on opposite thirds.
• Head and shoulders, not face-only. Back up until the frame holds the upper torso. It keeps gestures readable and avoids wide-angle distortion from sitting too close.
• No lines through the head. Check the background for poles, edges, or door frames crossing the head or neck. Move the camera, not the speaker.

2. Design the background

• Distance first. Get the speaker at least 4 feet off the background. Distance is what creates separation and soft blur; a flat wall 2 feet behind them will always read amateur.
• Blur beats clutter. When the background is busy, open the aperture (f/1.4 to f/2.8), lengthen the lens (85mm or longer), and increase the subject-to-background gap. If the camera cannot control aperture, distance alone does most of the work.
• Depth layers read as production value. Put something mid-distance between camera and background (a plant, a lamp, a shelf edge out of focus) so the frame has foreground, subject, and background planes.
• Slightly cooler background. Per the ASC's talking-head craft notes: a background a touch cooler than the subject reads as farther away (distant things look bluer), which adds perceived depth. Practical lamps stay warm on the subject.
• Backgrounds with leading lines and real depth beat flat walls, but only when you can keep them clean and lit sensibly. Remove or rearrange anything distracting before rolling.

3. Shoot with the edit in mind

• Two angles if you can. Main angle on axis, second angle 25 to 35 degrees off axis, tighter (main: headroom to mid-torso; cutaway: collar to tight head). Keep the speaker on the same thirds side in both so cuts feel continuous.
• Hold for punch-ins. Even one camera works if you frame slightly wide: the editor can punch in 110 to 120 percent at emphasis points. Plan emphasis lines during the read so you know where the punches go.
• Record clean air. Filler-free delivery matters less than a solid take; the edit will strip dead air, but it cannot fix a flat read.

4. Edit for retention

• Open on the best moment. The first 30 seconds has the steepest drop-off on YouTube. Start with the strongest claim, result, or problem, not a greeting or channel branding.
• Cut dead air ruthlessly. Pauses between sentences under 0.5 seconds, filler words and repeated sentences out. A rough cut should be 20 to 30 percent shorter than the raw footage.
• Never hold the same shot long. If the viewer watches one static talking head for more than 20 seconds, cut to something: B-roll, a second angle, an on-screen graphic, a text overlay for names and numbers. A practical floor: something changes visually every 10 to 20 seconds, calmly, not chaotically.
• Pattern interrupts every 60 to 90 seconds. A zoom cut, a graphic, a sound effect, a shot-angle change, a music shift. They break visual monotony and correlate with retention spikes.
• Use J-cuts and L-cuts to hide cuts. Let the next scene's audio start before the picture changes (J-cut), or let the current audio run over the next shot (L-cut). Micro-silences are where viewers leave.
• Punch in at emphasis, pull back at transitions. Quick punch-ins on key lines; ease back out when the topic changes. For revelations: slow push-in, then a beat of pause or a music drop before the line lands.
• Effects stay motivated. Every effect earns its place by serving the story. A stack of whooshes on a timer reads as cheap.

5. Finish pass

• Dialogue first: normalize speech around -14 LUFS integrated, keep music 15 to 20 dB below it, and duck the bed under speech.
• Match color across shots before chasing a cinematic grade.
• Add captions (burned-in or SRT); a large share of viewers watch on mute.
• Watch the full export start to finish before shipping. Every editor who skipped this has a story about the frozen frame at 7:42.

Output Contract

When asked to produce a talking-head plan or review, deliver:

1. Shot spec: aspect ratio, frame size class (medium / medium-close / close-up), speaker thirds position, eye-line direction, camera height.
2. Background plan: subject-to-background distance, depth layers, aperture/focal length if controllable, color temperature note.
3. Edit notes: planned punch-in points, B-roll slots with timestamps or script marks, interrupt placement (which type, where), music/ducking targets.
4. If given existing footage: a punch list of what violates sections 1 through 5, ordered by what the viewer notices first.

Operating Rules

1. Verify before asserting: any technical claim (a codec spec, a platform retention stat, a lens behavior) gets checked against a primary source. Cut what you cannot verify.
2. Calm over chaotic: retention editing adds change on purpose, never noise for its own sake. When in doubt, remove the effect.
3. The face is the anchor: framing, eye-line, and headroom decisions outrank background styling and color grade in every review.
4. Mobile viewers come first for vertical delivery: keep the speaker centered and the key action in the middle 80 percent of the frame so platform UI does not cover it.