In short: get the microphone close and the room soft; remove steady noise with AI; keep speech clearly louder than music (about 15–20 dB); cut the long pauses; and always check on phone speakers.
Why sound matters most
Our brains fill in missing detail in pictures but work hard to decode unclear speech, and that effort is tiring. Poor audio makes a speaker seem less credible — listeners rate the same talk as less intelligent when it's played with worse sound. So if you only have time to improve one thing, improve the voice.
Record it clean — the 80% fix
- Distance is everything. Halving the distance to the microphone makes the voice four times stronger relative to the room. A phone held at 30 cm beats a good camera microphone at 3 m. A cheap clip-on (lavalier) mic is the biggest upgrade you can buy.
- Soft rooms sound better. Bare walls and hard floors create echo that no software can fully remove. Record in a room with curtains, a rug, a sofa — or facing a wardrobe full of clothes.
- Turn off the hum. Fridges, fans and air conditioners are easy to stop hearing but obvious on a recording.
- Record 10 seconds of silence before you start. It helps you judge noise later.
- Don't clip. If the loudest parts distort, lower the input level or step back. Distortion can't be undone.
Remove background noise
Steady noise — fans, traffic rumble, electrical hiss — is what AI noise suppression handles best. VDOAPP uses a neural network (RNNoise) that was trained to separate speech from noise sound by sound.
- Select the clip and open the AI tab → Voice.
- Start with Noise removal strength at 100% and press 🧹 Remove background noise.
- Listen on headphones. If the voice sounds thin or "underwater", undo and try 70–80%: blending a little original sound back in keeps it natural.
In our test on a recording with a loud fan, the noise between words dropped from about −34 dB to effectively silent while the voice stayed intact. What noise suppression can't fix is echo (it's part of the voice) and other people talking (they are voices too).
Get the levels right
Uneven volume — one clip loud, the next quiet — is the most common problem in edited videos. Use the per-clip Volume slider in the Clip tab and the Mixer in the Audio tab to even things out:
- Match speech first. Play the cut between two clips of the same person. If the level jumps, adjust the quieter clip.
- Voice on top. Everything else — music, effects, ambience — sits under the voice.
- Fade edges. A Fade in / Fade out of a few tenths of a second on a clip's sound removes clicks at cuts.
Tighten the pauses
Natural conversation includes pauses that feel far longer on video. Remove silences in the AI tab finds them for you:
- Quieter than −38 dB is right for clean recordings; after noise removal you can go lower, around −45 dB, for more precise detection.
- Pauses longer than 0.6 s keeps a natural rhythm; 0.35 s gives the rapid YouTube style.
- Keep around speech 0.15 s protects breaths and word endings. If words sound clipped, raise it.
Mix music under speech
Music makes a video feel finished, but it's also the most common reason speech becomes hard to follow. A useful rule of thumb is to keep music roughly 15–20 dB below the voice — on VDOAPP's mixer, that's usually somewhere around 10–25% music volume depending on the song.
- Tick Lower other sound under voiceovers in the Audio tab so music dips automatically when a voiceover plays (called ducking).
- Choose instrumental tracks for speech. Lyrics compete with words for the same part of the brain.
- Avoid songs with lots of energy in the 1–4 kHz range (bright guitars, vocals) under speech — that's where consonants live.
Final checklist
- Listen once on headphones (for noise and clicks) and once on a phone speaker (for balance).
- Check the start and end — no abrupt sound at the first frame.
- Check the loudest moment doesn't distort after export.
- Watch once with captions on — mistakes in captions often reveal unclear speech.
Frequently asked questions
Can AI remove echo?
Not the noise removal in VDOAPP — echo is the room's reflection of the voice itself, so it's treated as speech. Prevent it when recording by using a soft room and a close microphone.
Should I remove noise before or after cutting?
Before captioning and silence removal, since both rely on the gaps between words being quiet. It doesn't matter whether you trim before or after.
Why does my voice sound fine on headphones but quiet on a phone?
Phone speakers can't reproduce low frequencies, so deep voices lose energy. Keep music lower and make sure speech is clearly the loudest element.