LEGAL / PRIVACY

Your words, handled with care

Comma collects the account, installation, usage and diagnostic information needed to provide the service. Recordings and history are stored on your device. Online recognition sends audio to a speech recognition provider.

Online recognition

Online recognition sends live audio and imported audio files directly from your device to Doubao Voice (Volcengine) for the China mainland App Store, or to AssemblyAI, with ElevenLabs as a backup, for other App Store regions and Android, which returns the transcript to your device. Comma receives session identifiers, audio duration, character counts, processing status and error codes. Comma does not receive or store the audio or transcript.

Where processing happens

Mainland China accounts, billing and usage data run on the mainland China deployment; international accounts, billing and usage data run in Silicon Valley, United States. The two business regions have independent accounts and balances. Apple payment notifications enter a shared verified routing service before regional processing. Doubao Voice processes domestic recognition through its China service. AssemblyAI and ElevenLabs processing locations depend on the service and data residency configuration available to the account. On-device recognition keeps the audio on your device.

Payments and subscriptions

We retain verified store transaction identifiers, product and billing-period details, subscription status, amounts, currencies and refund events to deliver purchases and reconcile your cloud hours. Credentials and full signed payloads are not exposed in public APIs or the admin detail view.

Usage and choices

Recognition and translation usage metrics are associated with an account or anonymous installation. On-device usage does not upload audio or transcript content. You can choose on-device recognition, submit feedback or sign out at any time.