Skip to content

Privacy

This is a beta build of a clip-processing tool. This page describes what the software stores so operators can adapt it. It is not legal advice, and a production deployment should have a policy reviewed for its jurisdiction.

What is stored

  • Account email, display name and a hashed password managed by the auth library.
  • Imported source video, a poster thumbnail, transcripts, clip metadata and rendered MP4s, in private object storage.
  • Processing events, usage ledger entries and error details needed to run and debug jobs.
  • Billing records: a payment-processor customer id, plan, subscription status and billing period. Card numbers are entered on the processor's hosted pages and never reach this application.

Retention

Sources are retained for 7 days and completed exports for 30 days by default. A scheduled cleanup job removes expired artifacts unless a job is still using them. Deleting a project revokes access immediately and schedules artifact removal. You can delete your entire account from Settings, which removes stored objects and database rows.

Third parties

Payments and subscriptions are handled by Stripe, which acts as the payment processor and receives the billing details you enter on its hosted checkout and customer portal.

When transcription or AI analysis is configured, extracted audio and/or transcript text is sent to the configured provider. When semantic vision is enabled, a bounded set of sampled JPEG frames is also sent to the vision provider so objects, visible faces and HUD/killfeed text can be observed. The full video is not uploaded. Without such configuration, those features stay disabled and the app says so. Demo mode uses a generated fixture and a deterministic transcript, clearly labelled in the interface.

Model training and providers

Customer content is not used to train or fine-tune models, and only providers on the deployment's allowlist may receive extracted audio, transcripts, prompts or sampled frames. A provider that reserves training rights is refused before any content is sent. The current allowlist and each provider's regions are published in the Data Processing Agreement.

Regional residency

Each deployment is pinned to a single region (EU or US). New workspaces are created in that region, and a storage or database region that contradicts the pin is treated as a configuration error instead of silently crossing the boundary. Multi-region routing is not offered; a second region is a separate deployment.

Limits

Sources are capped at 600 minutes / 10240 MB. These are configurable operational limits, not promises of capacity.

Terms · Back to home