TikTok Time

Talking-Head Video Playbook

You don't need word-for-word — you need a structural script: hook line, 2-3 beats, and how you're landing it. Talking-head lives or dies on knowing your next sentence before finishing the current one.
Write hooks to be spoken, not read. Say it out loud before filming — if it doesn't sound natural coming out of your mouth, it won't land on camera.
Pattern-interrupt/rage-bait hooks need delivery that matches the words. "Self improvement doesn't work" needs flat, deadpan, or slightly smug delivery to sell the reverse psychology — too earnest and it just reads as confusing.

Eye-level or slightly above camera. Below-eye-level is unflattering and reads as low-effort.
Rule of thirds: eyes in the top third of frame, not dead center — leaves room for on-screen text hooks.
Vertical 9:16 native. Don't film horizontal and crop.
Leave headroom, but not too much — excess dead space above the head looks amateur.

Front-facing soft light is non-negotiable. Cheap ring light or a window facing you (not behind you) fixes most "why does this look bad" problems.
Avoid overhead-only light (raccoon eyes) and backlighting (silhouette).
Match color temperature across lights — mismatched warm/cool looks unintentional and cheap.

Matters more than video quality. People tolerate mediocre video with clean audio, not the reverse.
Lav or shotgun mic beats built-in phone mic every time — even a $20 lav is a big upgrade.
Record in a room with soft surfaces (rugs, curtains, couch) — hard walls echo.

Energy up 10-15% from normal conversation — camera flattens energy, so what feels "too much" in the room reads as normal on screen.
Talk in complete thoughts, not paragraphs — easier to cut for pacing later.
Do multiple takes of the hook specifically — highest-leverage 1-3 seconds of the video, worth 5-10 takes.
Look directly into the lens, not the screen — breaks the "talking to a camera" feel, reads as eye contact with the viewer.
Natural hand gestures help — stiff hands-at-sides reads as nervous or scripted.

Cut on the breath, not the word — jump cuts should land where you'd naturally pause.
Jump cuts every 2-4 seconds keep pace up and remove dead space.
On-screen text hooks: separate layer from captions. Restate/amplify the spoken hook in fewer words, appear within the first second, bold/high-contrast font (white with black stroke or platform-native caption style).
B-roll/cutaways every 5-8 seconds for pattern-interrupt content — but for pure talking-head where the face IS the hook, don't over-interrupt; sustained eye contact can be the retention driver.
Zoom-punch (subtle scale-up cut) on key sentences to emphasize emotional or punchline beats.
Captions: auto-caption then manually clean up — auto-caption errors read as sloppy, especially on value-heavy content.

TikTok and Reels both still weight first-3-second retention heavily — hook-first is the right priority.
Native in-app filming sometimes gets a slight algorithmic nudge over uploads, though not consistent enough to rebuild workflow around — worth occasional testing.
Trend audio under talking-head can help discovery but shouldn't override the spoken hook — use as a low-volume background bed, not the driver.

Hook takes too long to say — front-load the punch, don't build to it
Flat/monotone delivery — even "controversial calm" hooks need vocal variation
Bad audio (echo, wind, low volume)
Too much dead space before getting to value
No visual movement for 60s straight — even subtle handheld sway or slow zoom keeps it from feeling stale