A voiceover is speech recorded by a speaker who is not seen on screen and is added to video, audio or interactive content to narrate, explain or persuade. In podcasts and speech production, the term covers both the performance itself and the recording and processing workflow that turns a read script into clean, intelligible audio.
Voiceover appears in documentaries, advertisements, explainer videos, e-learning courses, audiobooks, podcast intros and software prompts. Its quality depends on three things working together: a well-written script, a skilled read and a quiet, controlled recording that preserves speech intelligibility.
How a voiceover is produced
A voiceover is produced by recording a script in an acoustically treated space, editing the takes and processing the voice so it sits clearly under or over other audio. The workflow is usually linear, but each stage can expose problems from the previous one.
- Script preparation: the text is timed, marked for emphasis and checked for difficult names or terms.
- Recording: the speaker reads in a booth or quiet room, often with several takes per line.
- Editing: breaths, mouth clicks and mistakes are removed, and the best takes are assembled.
- Processing: equalization, compression and de-essing shape the voice for its destination.
- Delivery: the file is exported at the requested format and loudness, often with a separate stem for mixing.
In professional work, a director or client often listens remotely and gives line-by-line direction. This live feedback saves pickup sessions later, because tone and pacing are much harder to change once the speaker has left the booth.
Types of voiceover work
Voiceover work falls into several genres with different performance styles and technical expectations. A delivery that suits a car commercial would sound wrong in a medical training module, so speakers and producers match the read to the purpose.
| Genre | Typical length | Performance style | Pace |
|---|---|---|---|
| Commercial | 15 to 60 seconds | Energetic or warm, tightly timed | Varies with the brand |
| Documentary narration | Minutes to hours | Calm, authoritative, neutral | About 130 to 150 words per minute |
| E-learning and explainer | Short modules | Clear, friendly, instructional | About 140 to 160 words per minute |
| Audiobook | Several hours | Sustained, character voices possible | About 150 to 160 words per minute |
| Interactive and IVR prompts | Single phrases | Neutral, consistent across sessions | Moderate |
| Podcast intro and promo | 10 to 60 seconds | Conversational or branded | Moderate to fast |
Voiceover is distinct from dubbing, where an actor replaces on-screen dialogue in another language with matched lip movement. In translation voiceover, often used for interviews in news and documentaries, the original voice stays audible at low level while the translated voice speaks over it.
Technical requirements for a clean recording
A clean voiceover requires a low noise floor, short room reflections and consistent microphone technique. Because a voice is usually heard in isolation or with soft music, room sound and background noise are far more audible than in a dense music mix.
The table lists practical recording targets that many voiceover clients and audiobook platforms expect.
| Parameter | Typical target | Why it matters |
|---|---|---|
| Sample rate and bit depth | 48 kHz or 44.1 kHz, 24 bit | Matches video or audio delivery |
| Room noise floor | Below about -60 dBFS | Keeps pauses silent |
| Recording peaks | Around -12 to -6 dBFS | Leaves headroom for loud words |
| Microphone distance | About 10 to 25 cm | Balances presence and proximity effect |
| Room decay time | Very short, often under 0.3 s | Prevents a hollow, roomy sound |
| Delivered loudness | Set by the client or platform | Ensures consistent playback level |
A large-diaphragm condenser microphone is common in treated booths, while a dynamic microphone often works better in untreated rooms because it picks up less reflected sound. A pop filter and an angle slightly off the mouth axis reduce plosives on letters such as P and B.
Processing and mixing a voiceover
Voiceover processing aims to make speech clear, even and natural rather than impressive. Most chains are short, and every processor should solve a specific problem heard in the recording.
| Processor | Typical setting | Purpose |
|---|---|---|
| High-pass filter | About 70 to 100 Hz | Removes rumble and handling noise |
| Corrective EQ | Small cuts in boxy areas, often 200 to 500 Hz | Improves clarity |
| Compressor | Ratio 2:1 to 4:1, 3 to 6 dB reduction | Evens out loud and soft words |
| De-esser | Focused around 5 to 9 kHz | Softens harsh sibilance |
| Limiter or loudness stage | Ceiling around -1 dBTP | Meets delivery specifications |
When the voiceover is mixed with music, sidechain compression or careful level automation lowers the music under speech. The voice should remain at least several decibels above the bed, because speech intelligibility drops quickly when music competes in the 1 to 4 kHz range.
Common voiceover mistakes
The most frequent voiceover problems come from the room and the read, not from missing equipment. Processing can reduce a defect but rarely removes it completely.
- Untreated rooms add reflections that no plugin fully removes; blankets and close placement help more than an expensive microphone.
- Over-processing with heavy noise reduction creates watery artifacts that listeners notice at once.
- Inconsistent distance between takes changes tone and level, making edits audible.
- Reading instead of speaking produces a flat, announcer-like delivery that reduces trust in explanatory content.
A plain-language summary
A voiceover is a voice you hear without seeing the person speaking. It is the narrator in a nature film, the friendly guide in an online course or the announcer in a radio advertisement.
Making one well is mostly about a good script, a natural delivery and a quiet room. When all three are right, listeners stop noticing the recording and simply follow the message.
An example from practice
A company producing a twelve-part safety training course needs about 90 minutes of narration. The speaker records at home in a small closet lined with clothes, using a dynamic microphone about 15 cm from the mouth, at 48 kHz and 24 bit, with peaks near -10 dBFS.
The editor removes breaths and clicks, applies a high-pass filter at 80 Hz, gentle compression with about 4 dB of gain reduction and a de-esser, then sets every module to the same delivery loudness. Background music is lowered automatically under speech with sidechain compression.
Learners hear a consistent, clear voice across all twelve modules, and the training team receives no complaints about level jumps or unclear instructions.