Audio

How to Record a Clean Voice Over: Microphone, Room, Levels and Format

A good voice over is decided before the mic: in the room, the distance, the level. The rest is hard to fix later.

By Hanna Eng·Audio engineer, Abbey Road Institute Paris

Updated 15 October 20269 min read
Part of: Sound post-production

To record a clean voice over, treat the room first, place the mic second, and pick the model last. In an untreated room, use a cardioid dynamic mic, speak 15 to 20 cm away, slightly off-axis, with a pop filter. Record WAV at 48 kHz / 24-bit for video, with peaks between -18 and -12 dBFS, no compression or effects on the way in. And capture a few seconds of room tone for the cleanup.

A voice over carries a corporate video, a documentary, an ad, a tutorial. Audiences forgive average pictures; they tune out of a voice that echoes, pops on its p sounds or changes tone from one sentence to the next. Most of these problems are born at the recording, and no software removes them without a trace. I am a sound engineer: here is the order of priorities I follow, from the room to the delivered file, for a voice over that mixes without a fight.

Recording a voice over: starting settings

SettingRecommended value
Format for videoWAV 48 kHz / 24-bit
Format for an audiobook (ACX)Record at 24-bit, deliver MP3 192 kbps, 44.1 kHz
Recording level (peaks)-18 to -12 dBFS, never clipping
Distance to mic15 to 20 cm (6 to 8 in)
Angle15 to 30 degrees off-axis
Pop filter5 to 10 cm in front of the mic
Mic in an untreated roomCardioid dynamic
Processing on the way inNone: no compression, no EQ, no noise reduction
Room toneA few seconds of room silence, same setup

Source: Netflix dialogue recording specs (48 kHz / 24-bit); ACX submission requirements (official).

The room matters more than the mic

A €1,000 mic in a tiled kitchen mostly records the kitchen. The reverb of an empty room, with parallel walls and hard surfaces, sticks to the voice and never fully comes off in cleanup. It is the number one problem with voice overs recorded at home.

Pick the smallest, most furnished room: a bedroom with a bed, curtains, a wardrobe full of clothes. Clap your hands: if the sound lingers, the room rings. Thick blankets hung behind and around the mic, on the side you speak toward, do more for the voice than a new mic.

  • Avoid kitchens, bathrooms, empty rooms, large living rooms.
  • Shut off noise sources: ventilation, fridge, loud computer, notifications.
  • Put absorbent material behind you and in front of you, not only behind the mic.

Which microphone for voice over?

In an untreated room, a cardioid dynamic mic is the safest choice: it picks up mostly what is right in front of it and much less of the room. In a treated room or booth, a large-diaphragm condenser gives the voice more detail and air, but it hears everything else too.

In ADR and voice over for picture, you also see shotgun mics, very directional, which match well with production sound. A few neutral industry examples, with no commercial tie: Shure SM7B and Electro-Voice RE20 as dynamics, Neumann TLM 103 as a condenser, Sennheiser MKH 416 as a shotgun.

RoomMic typeWhy
Untreated, at homeCardioid dynamicRejects the room, forgives mistakes
Treated, boothLarge-diaphragm condenserMore detail, but hears everything
Matching production soundShotgunTone close to on-set mics

Distance, angle and plosives

Speak 15 to 20 cm (6 to 8 inches) from the mic, and hold that distance from start to finish: every time you lean in, the low end swells and the sound changes. Angle the mic slightly, 15 to 30 degrees, so your breath passes beside the capsule instead of hitting it.

A pop filter, 5 to 10 cm in front of the mic, tames p and b sounds. A simple marker: put your hand flat between your mouth and the mic to find the same distance at every take.

Recording format and level

For a voice over meant for video, record WAV at 48 kHz / 24-bit: it is the sample rate of audio for picture, the one required for example by Netflix's dialogue recording specs. For an audiobook, ACX delivery is MP3 at 44.1 kHz; recording at 24-bit is still good practice, the conversion happens at the end.

Set the gain so your loudest passages peak between -18 and -12 dBFS. At 24-bit, that headroom costs nothing in quality and protects you from a sudden loud line. A clipped take is lost: no tool truly repairs it.

  • Video: WAV 48 kHz / 24-bit.
  • ACX audiobook: record at 24-bit, deliver MP3 192 kbps, 44.1 kHz.
  • Peaks between -18 and -12 dBFS, never in the red.

Add nothing on the way in

No compressor, no EQ, no noise reduction while recording. Anything applied on the way in is baked into the file and cannot be removed. A raw, clean voice can be processed exactly as needed later; a voice already compressed and over-denoised cannot.

Monitor on closed-back headphones during the take: you will hear a mouth click, a creaking chair or a passing car right away, instead of discovering them in the edit.

Record room tone

Before or after the session, without moving the mic or the gain, record a few seconds of silence: the sound of the room with no voice. That is room tone. It does two jobs: filling gaps in the edit without the background changing abruptly, and giving the cleanup software a profile of the noise to remove.

Stay still and silent during that take. Room tone with a breath or a rustling page is useless.

The read: voice, pace, retakes

A voice over reads better standing or sitting upright, with the script at eye level so you do not drop your head toward a page. Drink room-temperature water and avoid dairy just before: it thickens saliva and multiplies mouth noise.

When you stumble, do not stop: pause, then restart the whole sentence from its beginning, in the same tone. In the edit, you just keep the good take. Picking up mid-sentence gives an intonation that never matches.

After the take: cleanup, mix, delivery

Once the voice is recorded, the order is always the same: edit the good takes, clean the problems one by one, then EQ, compress and set the loudness for the destination. For French TV, PAD delivery requires -23 LUFS and a -3 dBTP true peak; for the web, the target is louder.

For an ad, documentary or corporate film voice over, a session in a studio booked for the day, with an engineer at the desk, settles the room and level question in advance. That is also what I offer: the engineer behind the glass, not a studio for rent.

Frequently asked questions

What is the best microphone for voice over?

In an untreated room, a cardioid dynamic mic, which picks up little of the room. In a booth or treated room, a large-diaphragm condenser gives more detail. The mic matters less than the room and the distance.

How do I record voice over at home?

Pick the smallest, most furnished room, kill noise sources, hang blankets around you, speak 15 to 20 cm from a dynamic mic with a pop filter, and record WAV 48 kHz / 24-bit with peaks between -18 and -12 dBFS.

What format should a video voice over be?

WAV at 48 kHz and 24-bit, the standard for audio for picture. Keep the file raw, with no effects, and deliver it with a room tone file recorded in the same setup.

Should I use a compressor while recording?

No. Any processing applied on the way in is permanent. Record a raw voice with headroom; compression and EQ get set in the mix, when you hear the voice in context.

What is room tone?

A few seconds of silence recorded in the room, with the same mic and gain. It fills gaps in the edit without a change in background, and gives cleanup software a noise profile.

How do I avoid mouth noise in voice over?

Drink room-temperature water, avoid dairy and coffee just before, and keep the same distance to the mic. Whatever remains is removed in cleanup, one click at a time.

Sources and references

Have a project that needs this done right?

If your mix has to pass a platform spec, let's talk about the deliverables and the timeline.

Start a project