Asked twice, carried everywhere.
Silo Clone does not open until you confirm you will only clone your own voice or one you have explicit permission to use, and it asks again for every voice you build. That attestation is saved with the voice and written into every voice file you export, together with where the reference came from: a file, a link or a recording.
Hear it, scrub it, trim it.
Once the engines are installed, imported files and links have their backing music separated out and the voice levelled as they arrive. A recording gets the same treatment with one click. Each source opens as a scrubbable waveform: drag the in and out points and play just the selection. A voice keeps more than 5 and up to 15 seconds of speech, and Create Voice tells you when a selection is too short or too long. The readiness score follows your trim and says in plain words when a clip is too quiet, clipping or mostly silence, before you build a voice from it.
One file, wherever the voice goes.
Export Voice saves a .silovoice file wherever you choose. Silo Read, Silo Cast and Silo Fable open that file to speak in the voice. None of them records or builds a voice of its own, so the consent you gave here travels with the voice. Delete a voice in Silo Clone and its trimmed reference and every take are removed from this Mac; the recording it came from stays in the app's Studio until you delete it there.
Import or drop an audio or video file, paste a direct link to a media file, or record straight into the app from the input you pick, with a live level meter while you read.
High fidelity continues from your reference audio and its transcript for the closest match in timbre and delivery. Fast works from a compact voiceprint for quicker takes. Choose per take, star the ones you like, and share any take as an audio file.
Auto-transcribe writes down what the reference says, on your Mac, so High fidelity knows exactly which words it is hearing. Correct the text before you save the voice, or type it yourself when your language is not available for on-device transcription.
Pick input and output devices the way you would in Audio MIDI Setup, with each choice remembered, and choose which drive holds the 3.3 GB of engine files. Recording, importing, trimming and transcribing all work while that one-time download runs.
No account, analytics or telemetry, and nothing you record, import or generate is uploaded. The privacy policy lists the few downloads that do touch the network.