The short version
Silo Clone collects no data. None.
No analytics. No telemetry. No account. No personal data. No usage metrics. No error beacons. No advertising IDs. No fingerprinting. No crash reports. No data of any kind is collected, stored, transmitted, or shared by Silo Clone about you, your device, or your content.
Everything Silo Clone does — vocal isolation, levelling, transcription, voice cloning, and playback — runs on the Mac where you use it. There is no cloud inference, no server component, and no Silo Clone account.
Voices and recordings stay here
The audio you import, the vocal stems the app separates from it, the reference clips you trim, the voices you create, and every take you generate are written to disk on your Mac and never leave it. Silo Clone has no mechanism to send them anywhere. Sharing a voice or a take is something you do explicitly, to a destination you pick, using the system share sheet.
What we don't have, and therefore never see
Because Silo Clone has no server-side component, none of the following exists anywhere in our infrastructure (which consists of: nothing): your name, email, phone, or address; your IP address, device identifier, or any hardware fingerprint; your audio, your voices, your transcripts, or anything derived from them; session timing, feature usage, or behavioural data; crash reports. The app does not phone home, ever.
Outbound traffic you initiate
The only outbound network traffic, all user-initiated, and none of it to Silo Clone:
- Model weight downloads. Silo Clone offers to download the cloning engine from Hugging Face and the vocal isolation model from GitHub, both anonymously and only when you press Download. No token, account, or registration is required. Once the weights are on disk, Silo Clone works fully offline.
- A media link you paste. If you paste a direct audio or video file URL, the app fetches that file so it can be imported. It goes only to the address you typed. Streaming playlists and web page links are rejected; the app does not download from third-party media services.
- App Store updates. When Apple ships a Silo Clone update, the App Store delivers it. That traffic is between your device and Apple. Silo Clone is not involved.
- Links you click. If you click an external link inside the app, your default browser handles the request. Silo Clone is not involved.
Permissions we request, and why
- Microphone. Only to record a reference when you press Record. Nothing is captured otherwise, and recordings stay on your Mac.
- Speech recognition. To transcribe a reference on-device, so the high-fidelity clone knows what was said. Apple's on-device speech models are used; audio is not sent to a server for this.
- Files you select. To import the audio or video you choose, and to write exports where you ask for them.
Silo Clone does not request protected resources beyond those listed above. System permission controls remain in effect at all times.
On-device data you create
Sources, isolated stems, voices, and generated takes live inside the app's container at ~/Library/Containers/com.silosystems.siloclone/Data/Library/Application Support/SiloClone/. Model weights live there too unless you choose another folder in Settings. Files you import are read only for the workflow you started, and durable exports live wherever you save them. They are yours; we have no access to them. Removing the app bundle may leave sandbox data in place, so delete voices and models in-app first if you want them gone.
Sharing with other Silo apps on your Mac
If you publish a voice, it is written to a shared app group container (group.com.silosystems.voices) that other Silo and Ohms apps installed on the same Mac can read. That is a local folder on your machine, not a service. Nothing is uploaded, and publishing only happens when you switch it on for a specific voice.
Consent and provenance
Silo Clone requires you to affirm, at first launch and again for every voice you create, that the voice is your own or that you have explicit permission to clone it. That attestation is recorded alongside the voice on your Mac, and provenance metadata is embedded in voices you export. This record exists for your accountability, not ours: it is never transmitted to us or to anyone else.
Third-party components
Silo Clone uses the system frameworks, local model assets, and open-source components listed below. None receive any data about you from Silo Clone:
- Qwen3-TTS (mlx-community, Apache-2.0) — runs locally on-device.
- HTDemucs v4 (facebookresearch/demucs, MIT) — converted to Core ML, runs locally on-device.
- MLX and mlx-swift — local inference on Apple silicon.
- ZIPFoundation (MIT) — unpacks a downloaded model archive.
- Apple Speech, AVFoundation, Core ML, Core Audio — system frameworks, used on-device.
Children
Silo Clone is not directed at children and collects nothing from anyone, including children.
Changes
If this policy ever changes, the effective date at the top changes with it. A change that introduced any data collection would also require a new App Privacy declaration on the App Store, which is visible to you before you update.
Contact
Questions about this policy: admin@ohmslaw.net.