Support

You record a performance on your iPhone — video and audio together — or import something you already shot. An AI model running on the device splits that recording into two stems: the voice, and everything else. From there you balance the two, apply studio effects, correct pitch, add captions, master the result and export a finished video, a mixdown, or the separate stems.

It replaces the trip from your phone to a laptop and back for a large part of the work.

No. Recording, separation, mixing, tuning, captions, mastering and export all run on the device. The separation model ships inside the app, so there is nothing to download when you start a session and nothing to wait for on a server.

Airplane mode changes nothing about what the app can do. You need a connection only to download the app or an update from the App Store.

No. There is no server that receives your content, no account system, and no upload step anywhere in the app. Your recordings are never used to train or improve any model.

The only time your work leaves the device is when you export and share it yourself. The Privacy Policy sets this out in detail.

There is no account, no sign-in and no profile. Open the app and record. Nothing is tied to an identity because there is no identity to tie it to.

Yes. Vocals and instruments export as separate WAV files at 48 kHz, 24-bit, alongside the mixdown and the video. They open directly in Logic, Ableton, Pro Tools, FL Studio or anything else that reads WAV.

Yes. The licence covers commercial use, you own everything you make, and we take no share of it. The audio path runs at 48 kHz floating point end to end, and the preview engine and export engine are the same code — what you approve is what ships.

You are responsible for holding the rights to any material you bring in. Separating a commercial release into stems does not grant you the right to release the result.

They are deleted with it. Everything lives in the app's storage on your device, and we hold no copy that could be restored.

Export anything you care about, and keep your device backed up. If you have iOS device backup enabled, app data may be included in your own encrypted backup — that is a feature of your device, under your Apple Account, not something we control or can access.

Open iOS Settings → Take One and turn on Microphone and Camera. If the switches are missing entirely, delete and reinstall the app so iOS asks again.

Also check Settings → Screen Time → Content & Privacy Restrictions. If restrictions are on, camera and microphone access can be blocked there for every app, and the app's own settings page will not show it.

• Check that something is not covering the bottom microphone — a case, a grip, or a finger.
• If a Bluetooth headset was connected, iOS may have routed input to its low-quality microphone. Disconnect it and record again.
• Watch the input meter before you commit to a take. If Auto Gain is off, raise the input until peaks sit comfortably below the top of the meter.
• Another app holding the audio session — a call, a voice memo, a DAW — can capture the input first. Close it and restart Take One.

Clipping happens at the input and cannot be repaired afterwards — the peaks are simply gone. Leave Auto Gain on for loud material; it holds a safe level through a chorus. If you are setting gain manually, aim for peaks a little below the top of the meter and leave headroom for the loudest part you have not sung yet.

Distance helps more than settings: back off from the phone, or sing slightly across it rather than straight into it.

Incoming calls, alarms, low-power mode kicking in, and running out of storage will all end a take. Before a long session: turn on Do Not Disturb, check free storage, and start above about 30% battery.

4K at 60 fps consumes storage quickly. Dropping to 1080p roughly quarters the file size with no effect on audio quality.

Check iOS Settings → Take One → Photos. If it is set to Selected Photos, tap it and add the video you want, or switch to full access.

If the video lives only in iCloud and has not been downloaded to the device, open it in the Photos app first and let it download fully before importing — a partial file cannot be read.

A neural network trained on separated music estimates which part of the sound is a voice and which is everything else, then reconstructs each as its own audio track. It runs on the iPhone's Neural Engine and GPU.

Nothing is uploaded, nothing is queued on a server, and your audio is not used to train anything.

It depends on the length of the recording and the device — newer iPhones finish considerably faster. A typical few-minute take completes in a small number of minutes.

To speed it up: keep the app in the foreground with the screen on, close other heavy apps, take the phone off the charger if it has become hot, and avoid Low Power Mode, which throttles the processors the model depends on.

Separation is an estimate, not a surgical extraction, and quality depends heavily on the source. Results are hardest on dense mixes, heavy distortion, close vocal harmonies, loud room reflections and material where the voice and an instrument occupy the same range.

What helps most, in order:

• Record in a quieter, less reflective space — a soft-furnished room beats a tiled one.
• Get closer to the phone so the voice dominates the capture.
• Avoid clipping. Distorted input separates badly.
• If you are playing backing through a speaker, lower it relative to the voice.

Once separated, a small amount of Tone and the vocal fader will often disguise what remains better than trying to separate again.

Sung words are considerably harder to transcribe than speech — vowels stretch, consonants soften and pitch moves. Expect to review and correct captions rather than publish them untouched. Names, slang and code-switching between languages are the usual casualties.

Captions are generated on the device from your audio, so accuracy also improves when the vocal is clean and the backing is not overwhelming it.

That is expected. Running a separation model is one of the heaviest things a phone can be asked to do, and the heat is the cost of not sending your audio to someone else's data centre.

The app is tuned to stay within iOS thermal limits and will slow down rather than let the device overheat. Taking it out of a thick case, off the charger and out of direct sun all help it stay at full speed.

• Video — your performance with the finished audio, in MP4 (H.264) or HEVC, at the original frame rate or a fixed 30 or 60 fps.
• Mixdown — the finished audio on its own, as Studio WAV at 48 kHz / 24-bit, or a compressed high-quality file optimised for sharing.
• Stems — vocals and instruments as separate WAV files.
• Captions — burned into the video, or as a separate subtitle file.

Wherever you send them from the share sheet: your Photos library, the Files app, AirDrop to another device, or straight into another app. Nothing is saved outside the app unless you choose a destination.

If saving to Photos fails, check iOS Settings → Take One → Photos and allow adding to the library.

• Check free storage. An export needs room for the finished file in addition to the source material — for a long 4K take that can be several gigabytes.
• Keep the app in the foreground until it completes. Switching away can suspend the work.
• If it fails only with captions burned in, try exporting without them to narrow down the cause, then tell us — that is a bug worth reporting.
• Restart the phone and try once more before writing in.

Export the stems as WAV and save them to Files, then move them across with AirDrop, iCloud Drive, or by connecting the phone to your computer. They arrive at 48 kHz / 24-bit and line up from the same zero point, so they will sit in sync when dropped onto a timeline together.

The Master screen measures your mix and shows how it sits against common streaming references, with true-peak protection applied. Use it as a well-informed guide rather than a certification — always listen to the finished master before you release it, and check your distributor's own delivery requirements.