SOLIPSIUMAlpha

SOLIPSIUM Super-resolution audio engine

Engineer your perfect audio.

Bring a podcast, a video, a call or a voice memo. Solipsium will tell you what's wrong before it changes a thing.

Free for a limited time. No card.

Drop a recording here

It stays on your device until you ask for a repair.

MP3, WAV, M4A, MP4 and more.

Choose a recording
Hear a repair first

  • 90% of 1,344 speech recordings we damaged on purpose came out better. Measured September 29, 2026. Our strongest single tool managed 51.8%. How we measured it
  • No training on your recordings. And what you save is private to your account.
  • 30 new repairs a day on a free account. Whole recordings included.

A real voice, before and after.

We added the damage ourselves and matched the loudness. (A dropout is a short gap where the sound cuts out.)
0:00 / 0:08

Press play, then switch versions while it runs.

About this sample

Real speech with steady noise added, then cleaned with Natural speech cleanup. Both versions play at the same loudness. This is one example, and results depend on your recording.

Transcript: “It was served by sheriff officers, not by the council.”

VCTK 0.92, sample p225_044. Yamagishi, Veaux & MacDonald (2019). Source · CC BY 4.0. Noise, padding, cleanup, level matching and MP3 encoding added by Solipsium. Supplied without warranties; no endorsement is implied.

Try your own recording

How it works

It checks before it repairs.

One button, Suggested repair, runs all four steps.

  1. 1 Listens

    It names the damage first.

    You see each problem and where it is. Nothing has changed yet.

  2. 2 Repairs

    It repairs one problem at a time.

    Each one gets the repair that scored highest on that kind of damage in our tests, or gets left alone.

  3. 3 Labels

    It marks what it guessed.

    A filled gap is marked estimated. Rebuilt treble, inferred.

  4. 4 Returns

    You get the recording back.

    With a list of every change and its time. Your original is kept too.

Spectrogram of the damaged recording.Spectrogram of the recording handed back.
One sentence we damaged on purpose, then the same sentence after the repairs. The treble at the top of the second picture comes from AudioSR treble rebuilding, a separate step in the editor that Suggested repair doesn't run.
Hear this example and see how it was made

Or do it by hand. In the editor you pick the seconds that need work, then hear the repair against the original before you keep it.

A drawing of the editor. The Blend slider sets how much of the repair you keep.

What it fixes

Noise, echo, hum, clipping and dropouts.

Suggested repair picks from these, going by what the check found. In the editor, you pick.

Swipe sideways to see all three.

  1. Background noise and echo.

    A speech model lowers the noise around the voice and keeps the voice's full range. Echo has its own reducer, which takes a long recording about 20 seconds at a time.

  2. Hum.

    Solipsium finds mains hum at 50 or 60 Hz and its harmonics, then notches it out in bands a few hertz wide.

  3. Clipping and dropouts.

    Flattened peaks are partly raised back from the undamaged samples around them. Sharp-edged gaps of digital silence up to 100 ms are filled from both sides as estimates. No other web service page we read on October 3, 2026, names automatic dropout filling.

For muffled phone sound, hiss or crackle, try the free tools

Measured

We publish the misses too.

We measured Suggested repair on 1,344 damaged speech recordings Solipsium had never been tuned on.

90%

came out better, against 51.8% for our strongest single tool, run on every recording

That's 1,206 of 1,344, or 89.7%. They are 112 sentences from twelve speakers, each damaged 12 ways. We fixed the rules first, then opened the set once, on September 29, 2026.

3.1%

made worse, against 46.4% for that one tool

Each block is 78 dots. The red ones are the share of the 1,344 that came out worse.

This is real speech with damage we added, scored by a formula. Listeners haven't rated it. We haven't measured music this way, and Suggested repair has changed since September 29.

How it compares with other tools

Who it's for

Each kind has its own page.

Pick the one closest to your recording.

Podcasts, interviews and video.

The room was noisy, or your guest called in on a phone. Clean it up, in a sound file or under a video.

Clean up podcast audio

Calls and meetings.

A lost packet can leave a short gap of dead silence in a call. Solipsium fills those gaps, up to 100 ms each. It also turns down the fan behind a meeting.

Clean up a meeting recording

Film and documentary sound.

Location noise and hum on your production sound? Repair them here. Every change is listed by its time.

Solipsium for film

Voicemails and old tapes.

Repair just the seconds you can't make out in a saved voicemail or a tape transfer. The rest stays as it was.

Voicemails and old tapes

For business

Pilots for agencies, shops and archives.

Agencies, transfer shops and archives can start a pilot now. You get delivery checks, repairs you compare in your browser, and a report for each file. Pilots cost nothing during the alpha. The API isn't open yet. If you make laptops or PCs and want repair on the device, tell us what it would take.

Pilots and contactSolipsium for film

Our models

Prism

Our own models, each built for one kind of damage and measured before it touches your recording. The first is in the editor now.

  • The first one: lost audio

    Prism-PLC is for dropouts, the moments a call or a damaged file loses its sound. PLC stands for packet loss concealment.

  • When you get it

    Prism-PLC is in the editor now: an experiment members tick on per repair, built for speech. In an October 6 test, 75 of 80 recordings with dropouts measured better and none worse. On music it did worse than plain gap filling on 2 of 24 test clips. Next: a model for muffled speech.

  • How a model gets in

    A Prism model reaches the site only after it beats what it replaces on recordings it was not tuned on.

Questions

Before you upload.

Make a free account
What recordings can I work with?

WAV, FLAC, MP3, OGG, M4A, MP4, MOV, MKV and WebM, plus phone voicemails and voice memos in AMR, 3GP, AAC, Opus or WMA, up to 1 GB and 30 minutes in the editor. The check on this page reads up to five minutes. A project works on one audio track; you can pick a different track when you import.

Why not a free one-click cleanup?

Our strongest single tool, Full-band speech cleanup, made 46.4% of the 1,344 damaged test recordings worse when it ran on all of them (scored by formula). So Suggested repair checks first and runs only the repairs that scored highest for that kind of damage. What no repair improves is left alone, and you see every change before you keep it. We have not tested other products' one-click tools.

Does Solipsium change what was said?

No. The tools lower or remove sound, partly rebuild clipped peaks, notch out hum and fill short gaps, and none of them writes new words. A filled gap is labelled as estimated sound and rebuilt treble as inferred sound. A list called What changed points to moments where a repair added sound in a quiet place or removed loud sound, up to twelve of each, so you can listen there before you keep it.

What does Solipsium do with video?

It repairs the audio while you watch the picture in sync. A video export keeps the original picture with your repaired audio track and leaves out any other audio tracks and subtitles.

Do I need an account?

Not to try it. Drop a recording on this page: the check runs in your browser, and you can hear 20 seconds of it repaired without an account. To repair a whole recording and keep your repairs, make a free account. Your recordings, saved repairs and downloads are private to your account.

What happens to my recordings?

The quick check on this page stays in your browser. Repair the first 20 seconds sends those 20 seconds to the server, or the whole file, up to 25 MB, when your browser cannot open it or it runs over five minutes. The upload is deleted at once, and the 20 seconds and their repair within an hour; a try is never used for anything else. Files you import into the editor are uploaded to Solipsium's server and restricted to your account. Solipsium does not train models on your recordings. The privacy notice explains hosting, optional cloud processing, usage analytics and the limits on deletion. Keep your own originals, and upload only media you have permission to process.

Try it on your worst recording.

Make a free account and drop it in.

Repair a recording free

Nothing to install.