Support.
Written by the person who wrote the app.
That is expected on a new install. The voice engine and the isolation model are not part of the App Store download. Press Download Engines in the banner at the top of the window, or open Settings (Command-comma), Models, and press Download All. It fetches about 3.3 GB once. Recording, importing, trimming and transcribing all work while it runs.
By default inside the app's own container. Settings, Models, Download location shows the folder and the free space on that volume, and Change… lets you pick any folder you can reach, including one on an external drive. The choice is remembered across launches. Files you already downloaded are not moved: the new location starts empty, so download the engines again there.
Separating a voice from backing music is compute-heavy work that runs on your Mac's GPU, so a long import takes a while. The Studio shows how far it has got as a percentage. If isolation stops partway, you can still build a voice from the raw audio or try again.
Yes: Settings, Audio, Import processing. It is on by default because a reference built from a clean vocal clones better, but material that is already a dry voice does not need it. Recordings are never isolated automatically. Any source can be isolated by hand with Isolate Vocals in the Studio.
Check the readiness score on the Create Voice step. It flags clips that are too quiet, clipping or mostly silence, and it updates as you trim. Length is checked separately: a voice needs more than 5 and at most 15 seconds of speech, counted from the first sound. Create Voice opens on the first 15 seconds of a longer trim and tells you when a selection is too short or too long. Six to ten seconds of clean speech usually clones best. Then make sure the transcript matches the words exactly, because High fidelity mode relies on it.
Silo Clone accepts a direct link to an audio or video file only. Streaming playlists are refused before anything is fetched, and a web page link fails because there is no playable media behind it. Either way the app says why. It does not download from third-party media services.
Choose the input in the Record sheet before you press Start Recording, or set it in Settings, Audio, Input, which lists every input on your Mac with its channel count, the way Audio MIDI Setup does. If the chosen input is unplugged, recording uses the system default until it is back. Output has its own picker for playback.
Select the voice in the Voice Library and choose Export Voice (File menu, or Shift-Command-E). Save the .silovoice file wherever you like, then open it in Silo Read, Silo Cast or Silo Fable. The file carries your consent attestation and where the reference came from. Takes are shared separately, from the share button on each take.
In the Voice Library, choose Delete Voice from the voice's context menu or from its detail view. The voice, its trimmed reference and every take are removed from this Mac. The recording it was built from stays in the Studio until you delete it there, and voice files you already exported are not affected.
Macs with Apple silicon running macOS 26 or later. 16 GB of memory is recommended for the voice engine.
Include your macOS version, which Mac you are on, and what you were doing. There is no account to look up and no telemetry to consult, so what you tell me is all I have.