How to Clean Up Old and Archival Recordings
One sentence of scope before we start: we're fixing audio after it was recorded. If your noise problem is live, that's a different toolbox than the one on this page.
Archival cleanup has an order, and the order is the method: digitize at a healthy level → remove clicks and crackle → notch the hum → reduce hiss gently → EQ last. In Audacity that's Effect → Noise Removal and Repair → Click Removal, then the Notch Filter, then two gentle Noise Reduction passes (8–10 dB each) — and always on a copy, never the only original.
Verified in Audacity 3.7.8 on Jul 29, 2026 · applies to digitized tapes, voicemails, minidiscs, old phone videos · keep every original file, always.
Placeholder players — the real clips drop in at QA (see deploy notes for the file slots).
Step 0 — Digitize like an archivist
- Capture at a healthy level — peaks around −12 dB, never touching the top.
- Record the whole side, including the noise before and after the content; you'll want that noise for profiles.
- Save the raw capture as WAV and archive it untouched before any cleanup begins.
Why this works: every later step can be redone if the raw capture survives. The archival sin isn't bad cleanup — it's cleanup applied to the only copy.
Step 1 — Clicks and crackle first
- Run Effect → Noise Removal and Repair → Click Removal at the default threshold.
- Listen to a minute — if crackle remains, lower the threshold and repeat once.
- Fix stubborn single pops by hand: magnify the waveform and use Repair on selections under a few milliseconds.
Clicks confuse the noise profile you'll capture next — every impulse looks like signal. Clearing them first makes each later stage smarter.
Step 2 — Hum, then hiss, gently
- Notch the hum at 50/60 Hz plus audible harmonics (the hum & hiss guide has the exact settings).
- Capture a noise profile from the pre-content noise you deliberately kept.
- Reduce in two gentle passes — 8–10 dB each — rather than one heroic one.
- Finish with EQ: a gentle high shelf restores air the hiss passes spent.
When it fails: recordings where the voice itself is faint under heavy noise hit the floor fast. Accept the noise or escalate to specialist tools — more dB here only exhumes artifacts.
When the recording deserves the pro tool
- Know the ceiling: Audacity's chain rescues the common cases; badly degraded or historically important audio deserves iZotope RX, the restoration standard the pros reach for.
- Try the free chain first anyway — you'll learn what the recording needs before paying for it.
- For voices that matter and budgets that don't stretch, Adobe's Enhance can resurrect intelligibility, at the price of its produced tone.
Restoration is triage: most files need the free chain, a few need the specialist, and knowing which is which is the actual skill this page teaches.
Archival audio punishes over-suppression hardest because there's so little signal to spare: one aggressive pass and grandma's cassette letter becomes a lisping voice underwater, with musical-noise sparkles where the room used to be. The fragile-material rule is halve whatever setting you'd use on modern audio, run twice as many gentle passes, and A/B against the raw capture at every stage — the raw file you archived in Step 0 is the conscience of the whole project.
Still noisy? Escalate in this order
- Voice barely there under the noise? You've met the floor — a specialist tool or a transcription is the honest endpoint.
- Speed or pitch sounds wrong? Fix that before any cleanup — tape speed errors masquerade as tonal problems.
- Only one playback of a dying tape left in it? Capture raw, touch nothing, and consult an archive service before experimenting.
Quick questions
What's the best free tool for cleaning old tape recordings?
Audacity, used in the archival order — clicks, then hum, then gentle hiss passes. Its Click Removal plus the two-pass noise method covers most home archives; the discipline matters more than the software (checked against 3.7.8, Jul 29, 2026).
Should I use AI enhancement on historical recordings?
Carefully. Speech reconstruction like Adobe's rebuilds the voice, which restores intelligibility but replaces some of the original character — an archival trade-off, not a free lunch. Keep the raw capture; enhancement is a rendition, not a restoration.
What does iZotope RX do that Audacity can't?
Spectral repair — editing individual sounds out of the audio picture, de-clipping, de-reverb, and machine-learning modules tuned for restoration. It's the professional standard; the free chain here is the right first mile before deciding you need it.