Audio restoration for vinyl, cassette and tape

Tool for restoring old audio in your browser. Diagnose noise, clicks and bandwidth automatically, then apply a per-source preset for vinyl, cassette, 78 RPM, shellac, reel-to-reel or voice archive. Paint over unwanted artefacts directly on the spectrogram, compare original to restored side by side, and export to WAV, FLAC, MP3 or OGG with full bit-depth preservation.

1
Upload your recording

WAV, FLAC, MP3, OGG, M4A or AAC. Bit-depth is preserved through the chain.

2
Read the diagnosis

The diagnostic card highlights what needs attention and recommends a preset.

3
Pick a preset or build a chain

Choose a source preset or open Custom for full manual control of every module.

4
Compare before and after

A/B toggle with synced spectrograms and LUFS readout. Iterate until the result is right.

5
Repair the spectrogram (optional)

Open Spectral repair and paint over remaining bursts the chain did not catch.

Clean clicks, crackle and hiss off a digitised recording

0:00
Drop a file here
WAV, FLAC, MP3, OGG, M4A, AAC ·
🎞️

What is your source?

Diagnosis Analyzing recording…

Overload flat-topped peaks — captured too hot
Noise and hiss the steady background of tape or preamp
Pick a silent region
Noise profile auto

Drag on the waveform to mark a region without music or voice — that becomes the noise reference.

Clicks and crackle scratches and surface noise of a disc
Decrackle

Smooth sustained surface crackle

Mains hum 50 or 60 Hz and its harmonics
Channels a mono disc read by a stereo cartridge
Bandwidth above this the medium stored nothing
Fine tuning

Processing order drag to rearrange

Peak reconstruction Fold to mono De-rumble De-hum Declick Decrackle Dehiss De-emphasis EQ correction Loudness target
De-rumble

Cut subsonic noise below the cutoff

Dehiss
De-emphasis

Apply inverse-RIAA / 78 / CD curve

EQ correction

4-band parametric equalizer

Loudness target

Normalize to EBU R128 broadcast level

Why this recommendation?
{}
Restoring…
5-second preview ready

Spectral repair

Paint over unwanted bursts on the spectrogram. The selected time × frequency cells are replaced with the local-median magnitude — without phase artefacts.

1.0×

Pinch to zoom, single-finger to paint.

Queue

Drop more files to add to the queue

💡 Have more files like this? Switch to Batch mode to apply the same chain to all of them.
Don't know what to try?
700+ tools — open a surprise
🎲 Surprise me

Features

Automatic diagnosis
Measures noise floor, click density, mains hum, bandwidth and loudness, then recommends a preset within 150 ms.
Seven source presets
Vinyl LP/45, 78 RPM, cassette, reel-to-reel, shellac, voice archive, plus full manual control.
Re-orderable DSP chain
Eight modules — de-rumble, de-hum, declick, decrackle, dehiss, de-emphasis, EQ, loudness — each independently toggleable.
Learn-from-selection dehiss
Mark a silent region on the waveform; the noise profile is learned and subtracted spectrally.
Spectral repair canvas
Paint rectangles or freehand over the spectrogram to remove specific time × frequency artefacts.
EBU R128 loudness target
Two-pass loudness normalisation to −23, −16 or −14 LUFS for broadcast, podcast or streaming.

What sources are supported?

Vinyl LP, 45 and 78 RPM, compact cassette, reel-to-reel tape, shellac/wax cylinder transfers, dictaphone and voice-memo archives. The custom preset works on anything.

What is the difference between declick and decrackle?

Declick repairs short transient clicks — single-sample to 5 ms pops typical of vinyl scratches. Decrackle smooths sustained surface noise — the continuous low-level crackle of a worn record.

How does learn-from-selection dehiss work?

You drag a small region on the waveform that contains only noise — no music or voice. The tool measures the noise spectrum there and subtracts it from the whole recording, preserving signal that rises above the noise floor.

Can I restore a 78 RPM record?

Yes. The 78 RPM preset applies aggressive declick and decrackle, band-limits to 100 Hz – 4 kHz where the signal lives, and offers inverse curves for NAB, Columbia and RCA pre-emphasis.

What is the spectral repair canvas?

It shows the recording as a spectrogram. You paint rectangles or freehand over short unwanted events — a cough, a microphone bump, a specific dropout. The selected time × frequency cells are replaced with the local-median magnitude, leaving phase coherence intact.

Will this damage a clean modern recording?

No. The diagnosis flags clean recordings, leaves all modules off by default, and the chain is a no-op when no module is enabled. A specific negative-control test verifies that processing a clean 16/48 file changes loudness by under 0.5 dB.

Which formats and bit depths are supported?

Input: WAV, FLAC, MP3, OGG, M4A, AAC. Output: WAV (16/24/32-bit), FLAC (16/24-bit), MP3 (up to 320 kbps), OGG (up to q10). Source bit-depth is preserved end-to-end where possible.

Why might the diagnosis differ from my own?

Diagnostic heuristics analyse an 8-second window from the middle of the file. If your source has unusual structure — silent intro, a single loud event, or atypical hum frequency — the heuristic may pick a different preset. You can always override or pick Custom.

💡 Want us to improve this tool just for you?

We can! Just send us a quick message with your idea. If you'd like to discuss it in detail, leave your email and we'll get back to you. You can stay anonymous.

How do you rate this tool?

Thank you for your rating!
Want to share more? Leave a comment!
Thank you! Your comment will appear after moderation.
Published Updated Author: