What is a talking head video?

One person, one lens

The format is defined by what is missing: no set, no action, no second speaker, just someone telling you something. In production terms it is A-roll, which Adobe describes as footage featuring key actions, talking characters or interview subjects, as opposed to the supporting shots laid over it. There are two versions. Direct address has the speaker looking into the lens and talking to the viewer, which is the norm for vlogs, lessons and ads. The interview version has them looking slightly off camera at an unseen interviewer, which reads as more documentary and less like a pitch.

How it is shot and cut

Framing sits between a medium shot and a close up, with the eyes about a third of the way down the frame. Sound matters more than picture: viewers forgive a soft image but leave over echo or a distant voice, so a small microphone near the speaker is the first purchase. Soft light from the front or side keeps the face readable. In the edit, pauses, stumbles and retakes are removed with cuts, which is where the familiar jump in framing comes from, and supporting shots are laid over some cuts so the join is hidden and the viewer sees what is being described.

What the research found

Two studies are often quoted, and they only seem to disagree. Jakob Nielsen's 2005 eyetracking work concluded that a talking head is boring online even for 24 seconds, with attention wandering to objects in the background, and advised keeping web video under a minute. Philip Guo, Juho Kim and Rob Rubin's 2014 study of 6.9 million viewing sessions on four edX courses found shorter videos far more engaging and informal talking head footage more engaging too. Read together, the lesson is that a face on its own is weak, while a short, informal, personal delivery holds attention well.

When a face matters

A talking head earns its place when the viewer is deciding about a person as much as a product: a consultant, a coach, a clinic, a founder asking for trust after a problem. It also suits expertise, where the credibility of the explanation depends on who is giving it, and community content, where the regular presenter is the reason people return. For a product that is visual, a process with steps or a claim that needs numbers, the face is better used briefly at the start and end, with the rest of the video showing the thing itself. It is also the quickest format to publish against a deadline, since someone who knows the subject can record a clear update in one sitting.

Common mistakes

Reading visibly from a script is the most common, since viewers see the eyes track across a line. Recording one long take and publishing it with the pauses left in is next. Poor audio loses more viewers than poor lighting. A busy or distracting background competes for the attention Nielsen found wandering. Leaving out captions ignores everyone watching muted in a feed. And spending the first ten seconds on a greeting and a channel introduction delays the only thing the viewer came for, which is the point of the video.

Where it shows up in the product

The video builder here does not film anyone. Its closest format is the UGC video type, on the Pro plan or higher, where the spine of the video is up to 3 talking scenes spoken by a generated creator: one Gemini portrait, animated by WaveSpeed's Kling avatar model, with one synthesised voice for every line, and quick proof beats cut between the talking scenes. The whole video stays under 30 seconds. That is a synthetic talking head, not a recording of you. For a real one, film it yourself and add it to a promotional or tutorial video as a clip scene: MP4, WebM or MOV up to 200 MB, with trim, cover or contain fit, and the clip's audio kept or muted. Background music does not duck under a clip's own sound, so lower the music whenever the clip carries speech.

Questions people ask