New in 3.8: room echo and mouth clicks — one knob
- De-reverb (room echo) — recorded in an empty room and it sounds like a bathroom? The late reverb tail is reduced while timbre and clarity are preserved.
- Mouth de-click — lip and saliva clicks in pauses are suppressed automatically; consonants stay intact.
- The 'Voice / podcast' mode — one click enables de-reverb, mouth de-click, gentle vocal denoise and de-esser together. Buttons on this page open the app with the mode already on.
Episode preparation checklist
- Quiet room or denoiser — record without background noise or remove it in post.
- Balanced speaker levels — equalize loudness so listeners don't adjust volume.
- Sibilance under control (de-esser) — tame harsh 's' and 'sh' sounds for headphone listeners.
- Target LUFS — Apple Podcasts recommends −16 LKFS ±1 dB. Other platforms' targets are in the table below.
- True Peak ≤ −1 dBTP — prevents clipping during AAC/Ogg transcoding.
Questions and answers
What loudness does a podcast need?
The standard for calm speech is −16 LUFS integrated (the Apple Podcasts recommendation) with true peak at or below −1 dBTP. That is quieter than music at −14 LUFS so the voice stays even and easy on the ear. The «Speech» preset in Magic Master levels a podcast to exactly −16 LUFS.
How do I remove room hum and noise from a recording?
The «Speech» preset automatically applies a rumble filter (cutting low-frequency rumble and air-conditioner hum) and gentle noise reduction, and lifts intelligibility in the 2–5 kHz range. No reverb is added — the voice stays close and clean.
Does it work for a remote interview or a phone recording?
Yes. We clean up steady noise and rumble and even out levels so that different sources (a guest on a call, a phone recording) sit together better. For difficult cases with heavy echo the result is better from a recording made in a quiet room.
How much does podcast processing cost?
The first basic mastering each day is free (by UTC), no card needed. Mastering with all PRO modules is 1 token = 1 episode, paid in roubles via YooKassa.
What format should I upload: WAV, FLAC or MP3?
WAV or FLAC. Every lossy encode adds artifacts to speech, and an MP3 input cannot remove them later. Keep an uncompressed file through all editing and make the MP3 the very last step, for publishing.
Should I level each line separately?
No. Loudness is judged over the whole episode. Levelling each line on its own removes the natural shifts between speakers and laughter, and the overall level drifts away from the target. Balance the speaker tracks against each other once, then process the episode as a whole.
What if the recording is clipped?
Neither loudness nor a limiter brings back clipped peaks: they are already lost. If you can, re-record the passage or use a second take. The separate de-clipper tool (/en/declipper) can partly repair the damage, but it cannot fully restore the original waveform.
Should intro music and stingers be processed together with the speech?
Better separately. Music has different dynamics and its own limits, and processing it should not pull the voice down. Assemble the episode from finished files, then check loudness on the final file.
See also: Mastering for VK Music, Yandex Music & Zvuk · Telegram bot · Pricing & features