How to Edit Multi-Track Interviews Fast With Cleanvoice
Handle multi-track interviews in Cleanvoice by uploading each speaker as a separate file, enabling multitrack processing, and cleaning at conservative strengths so crosstalk isn’t destroyed. A two-person, 60-minute episode costs a few dollars and about 30 minutes of your time.
The single-track workflow breaks down at two speakers. Cuts land in the middle of the other person’s sentence, overlap gets clipped, and the filler removal that looked fine on a solo recording starts eating real words. Cleanvoice has a multitrack mode that mostly solves this, provided you recorded in a way that lets it work.
What does multitrack mode change?
Each track is analyzed separately, so a pause is detected against that speaker’s own audio rather than the mixed signal. When one person talks over another, the tool doesn’t cut the quieter speaker’s mid-word.
It requires separate files per speaker. If you recorded a mixed stereo file, you can’t retroactively turn it into multitrack — a hard constraint people discover after recording, which is the expensive way to learn it.
How do you prepare tracks for upload?
- Export one WAV per speaker, 48kHz, no processing.
- Name them by speaker, not by device.
- Note the sync offset if the recorders clocked separately; Cleanvoice aligns tracks, but a manual offset saved before upload avoids surprises.
- Make sure no track is clipped — check your DAW’s peak meter before exporting.
- Trim leading silence so both tracks start at roughly the same timecode.
Step 5 isn’t strictly required but shortens processing and reduces sync errors.
What’s the editing sequence that works?
Clean first, edit second. Not the other way around.
- Run multitrack cleaning with conservative settings.
- Pull the cleaned tracks into your editor and review the joins between speakers.
- Remove content — tangents, false starts, repeated stories.
- Level-match and add music.
If you cut content first and clean afterward, you pay for minutes you’ll delete and you clean audio you’re not using. If you clean too aggressively first, you’re cutting content out of already-damaged audio.
How do you review for over-editing?
Listen at 1.5x speed. Problems are more audible when sped up — phrases that lost a syllable sound clipped, and unnatural silences stand out.
Three checks, five minutes total:
- Speaker handoffs: does each reply start cleanly, or is the first word missing?
- Laughter and pauses: are natural beats intact?
- Numbers and names: did aggressive filtering eat a consonant at the start of a name?
If a section sounds wrong, don’t re-run the whole episode. Re-run only that file, or fix the section by hand.
| Episode type | Tracks | Typical processing time |
|---|---|---|
| Solo, 30 min | 1 | 6-10 min |
| Two-person, 60 min | 2 | 15-25 min |
| Three-person, 90 min | 3 | 25-40 min |
Where does the tool stop being enough?
When the conversation overlaps heavily. Multitrack protects against cutting mid-word, but it can’t separate two voices recorded on the same microphones. Roundtable shows with shared mics need source separation first.
It also can’t handle level problems well. Cleanvoice does some normalization but not broadcast-grade loudness matching. If speaker A is 8dB hotter than speaker B, fix it before upload.
Finally, there’s no transcript-based editing. You edit blind, by ear. For shows where you cut heavily, an editor with text-based editing will still be faster overall.
Who fits this pipeline?
Two-person interview shows with separate mics, weekly publishing schedules, and hosts who’d rather do one review pass than scrub a waveform.
Not for roundtable recordings on shared mics, not for shows with heavy content restructuring, and not for anyone who needs loudness-compliant masters straight out of the box.
Quick Summary
- Multitrack mode processes each speaker separately so crosstalk isn’t cut mid-word.
- You need separate WAV files per speaker; mixed stereo can’t be converted later.
- Clean first, then cut content, then level-match — the reverse order wastes credits.
- Review at 1.5x speed and check speaker handoffs for clipped first words.
- Two-person 60-minute episodes run a few dollars and 15-25 minutes of processing.
