Voiceover is a second author arguing with the picture

Everything below about voiceover narration comes from what actually happens rather than from what is supposed to.

A narrator is a character even when unnamed

The moment a voice comments on the images, the audience begins asking who is speaking, when they are speaking from and what they want the listener to think. Those questions arrive whether or not the film intends them, which is why an anonymous explanatory voice still carries an implied position.

A narrator located after the events has already survived them, which quietly removes a category of suspense the film might have wanted. A narrator speaking in the present tense preserves that uncertainty but has to explain how the commentary is being produced. Choosing between those positions is a structural decision with consequences across the whole running time.

The voice and the image can disagree

The most productive use of narration is contradiction, where what is described and what is shown do not match and the audience has to decide which to trust. This is the mechanism behind unreliable narration on screen, and it depends entirely on the camera retaining independent authority.

If the pictures merely illustrate the words, the film has spent a resource and gained nothing beyond a slightly faster delivery of information. Comedy uses the same gap constantly, since a confident description undercut by a visible reality is one of the sturdiest jokes available. Documentary practice has argued about this for decades, with the authoritative unseen commentator contested precisely because it suppresses that disagreement.

Narration written late is usually a repair

When an audience cannot follow a story, adding a voice is the cheapest available fix because it requires no reshooting and no restructuring. The symptom is narration that appears heavily at the start, disappears through the middle and returns to explain the ending.

By the second act, narration designed from the beginning behaves differently, threading consistently through the film because it is part of how the story is told rather than a patch. This distinction is visible without any knowledge of a production history, simply from where the voice appears and what it is doing when it does. Films have been substantially altered by such additions in both directions, and versions with and without narration sometimes circulate separately.

What narration is genuinely good at

Interiority is the clearest case, since a character’s private assessment of a situation is otherwise expensive to dramatise and easy to get wrong. Compression is the second, because a sentence can cover years that would otherwise consume a sequence the film cannot afford. Voice also establishes register immediately, telling an audience within seconds whether they are watching something wry, elegiac or clinical.

Adaptations use it to retain a prose voice that carried much of the source, which is a legitimate choice rather than a failure of nerve.

In each case the voice is doing something the image cannot, which is the only durable justification for using it.

What it is bad at

Describing emotion that is already visible on a face is the most common misuse, and it flattens the performance rather than supporting it. Explaining plot mechanics tends to signal that the mechanics were unclear, and audiences read the explanation as an admission. Continuous narration across a whole film risks converting cinema into illustrated audio, where nothing is discovered by watching.

Excessive irony wears out quickly, since a voice that undercuts everything eventually undercuts the moments the film needs taken seriously. A useful discipline is removing every narration line whose information is available in the frame and seeing what genuinely needs to return.

Release strategies shift between territories, so your experience may differ.

Recording and placement change the effect

Narration recorded close and dry sits inside the listener’s head, while a version with room around it sits in a place and implies an audience. Where a line begins relative to a cut alters its meaning, because a voice that starts over the outgoing scene comments on it rather than on the new one. Level relative to music and effects determines whether narration feels like thought or like address, and mixers treat that as a dramatic choice.

What the cut is doing: performers deliver narration differently from dialogue, generally quieter and slower, because the microphone is closer and the intention is confiding. These are not technical afterthoughts; they decide whether the voice belongs to the film or hovers above it.

The takeaway

Ask what the voice knows that the camera does not. If the answer is nothing, it is a patch.

Craft is the part that keeps working after the surprise has gone.