SOLIPSIUMAlpha

SOLIPSIUM Super-resolution audio engine

Voice isolator

A voice isolator for speech recordings.

Solipsium lowers the noise and chatter around a speaking voice; it does not separate vocals from music.

Drop a spoken recording here

Solipsium checks it in your browser for clipping, hum and noise, without uploading it. It can then repair 20 seconds for you to hear, with no account.

WAV, MP3, M4A, MP4 and more, up to five minutes for the check.

Choose a recording

What it is

The voice in front stays.

Solipsium repairs noisy or damaged recordings. When it hears other voices or steady noise behind a speaker, it runs Full-band speech cleanup. That speech model keeps the voice's full range and lowers what surrounds it.

  1. The check runs in your browser.

    It names background noise, mains hum, clipping, dropouts and muffled sound. Other voices are measured on Solipsium's server, once you ask for the try.

  2. The try repairs 20 seconds.

    Solipsium's server measures those 20 seconds and lowers what it finds behind the speaker. Then switch between Original and Repaired as they play.

  3. The editor takes the whole file.

    With a free account, it repairs up to 30 minutes at a time and adds tools of its own, such as Rebalance for music under a voice.

How it is repaired

Chatter and noise come down.

Suggested repair picks one cleanup from what it measured, and names it.

Other voices are lowered.

A café or a television can sit behind the voice. When the try hears one, it lists the reason: “Other voices seem to sit behind the speaker, so speech cleanup is used.”

Steady noise gets the same cleanup.

A fan or traffic within 26 dB of the voice is lowered by the same speech model.

Two close voices were not measured.

When two people are equally close to the microphone, switch between Original and Repaired on the same words before you keep the result.

Questions

What it is not for.

Can it take the vocals out of a song?

No. Solipsium's free try is built for speech, and it has no step that separates a singer from the music. The editor's Rebalance is meant for a voice over a bed of sound, such as talk over background music.

Can it turn down music under a voice?

In the editor, with a free account: Rebalance finds the voice with a speech model and turns the rest down. In an October 5 test, the music usually went down by less than the setting asked. At the default setting the voice measured worse than before in 39 of 317 test recordings, 38 of them with a made-up chord as the music.

Will my own voice change?

It can. The editor's note on Full-band speech cleanup asks you to check quiet or distant voices, so switch between Original and Repaired on the same words before you keep it.

Can I choose which voice to keep?

No. The repair keeps the voice in front and lowers the rest. A recording of each person on their own track keeps them apart.

Is my recording uploaded?

Not for the check, which runs in your browser. Pressing Repair the first 20 seconds sends those 20 seconds. If your browser could not open the file, or it runs over five minutes, the whole file goes to Solipsium's server instead, so the try is offered for such a file only up to 25 MB. Solipsium keeps only those 20 seconds and deletes them within an hour. If you are signed in, a file you drop or choose is uploaded to your account and opens in the editor instead. The privacy notice has the rest.

Free audio tools

Related free tools.

Each of these takes one kind of sound behind a voice.

Want Solipsium to do the whole recording the same way?

The editor runs the same Suggested repair on the whole file, up to 30 minutes at a time.

Repair the whole recording

Free for a limited time, whole recordings included. No card.