A surprising amount of video work is not talking to camera but talking under it: narration over a screen recording, a voiceover for b-roll, the explainer where the visuals carry the picture and your voice carries the argument. The instrument is the same phone, and most of the craft transfers untouched — minus the light, which no longer matters at all. What remains is sound and delivery. A microphone a hand-span from your mouth. A read paced for someone listening, which is slower than feels natural when reading. And a script in front of you that does not rustle like paper or wander like a half-remembered outline. A voiceover recorded as an afterthought, into a phone lying on a desk across the room, is the most common way an otherwise well-made video announces what it cost.
The room matters more than the microphone. Hard parallel surfaces give phone recordings that boxy, echoey sound, and no setting removes it afterwards. Soft things fix it: face a duvet, record inside a wardrobe of hanging clothes, pick the smallest carpeted room and get close to the soft surfaces and away from the glass. Then hold your distance. A hand-span from mouth to microphone, kept constant, does more than the choice of microphone ever will — drifting nearer and further between sentences is audible as the voice thickening and thinning. If your consonants pop, speak across the microphone rather than into it. None of this costs anything, and all of it carries into every voiceover you record afterwards.
Talk2Camera records audio-only takes: just the sound, with the same prompter in front of you and the same pause behaviour, so paused time is excluded from the file rather than recorded as dead air. When the narration is done, attach it to the edit as an audio track, positioned where you want it — preview and export both honour the placement. The take’s own sound and the attached track each carry a level line you drag across the waveform, the way a mixer works rather than a fader, and the bars redraw at the level you set — music pulled down under narration looks quieter as well as sounding it. Laying a bed under a voiceover is one drag, not a project.
The raw audio never leaves the phone. A take is a file on your device, not a session on somebody’s server: the recognition behind the prompter’s voice follow runs on the device, and the only thing that leaves is an export you choose to share. For narration this matters more often than people expect, because voiceover work is so often about things that are not public yet — a product that has not shipped, a client’s course, a video essay over footage licensed on conditions. It also removes a practical tax: there is no upload to wait for. The moment you stop reading, the take is sitting next to the edit, ready to be laid underneath it.