What data we collect

Account data (only if you sign in)

• Your email address (from Google OAuth).
• Your Google account ID (an opaque string we use to match you to your sessions).
• Your display name and avatar URL (used to show 'Hi, [name]' in the header).
• The date you signed up.
• Your tier (free / Pro) and Stripe customer ID if you've paid.

What payment data does AetherCut collect?

AetherCut stores Stripe payment intent ID, session ID, and customer ID, plus purchase amount, currency, timestamp, and refund or dispute status. Stripe handles card details directly, so AetherCut does not receive your card number.

• Stripe payment_intent_id, session_id, and customer_id. We never see your card number — Stripe handles that directly.
• Purchase amount, currency, and timestamp.
• Refund / dispute status (so we can revoke Pro access on refund — see /api/webhook/stripe).

Does AetherCut store my IP address or track me across sites?

AetherCut does not store your full IP address and does not set cross site tracking cookies. It logs visited paths, referrer host, UTM tags, device type bucket, and timestamps, without joining page views to your account.

• The path you visited (/, /security-video- editor, /about, etc.).
• The referrer host (e.g. google.com, news.ycombinator.com).
• UTM tags from the URL, if present.
• User-agent bucket (desktop or mobile — not the full UA string).
• Timestamp.

We do NOT store the full IP address. We do NOT set cross-site tracking cookies. The page-view row is not joined to your account record.

What happens to my data when I use AI features?

When you use an AI feature, the relevant input is sent to a third party model, but AetherCut keeps only the output. For auto captions, audio is deleted within seconds, and Privacy Mode disables these features.

When you trigger an AI feature, the relevant input is sent to a third-party model. We retain only the OUTPUT (e.g. the generated subtitle text), not the input.

• Auto-captions: extracted audio → Whisper → text result. The audio file is deleted within seconds. The text is saved to your project if you keep it.
• AI Director: timeline metadata → Claude → suggestion text.
• AI B-roll: text prompt → Kling / Nano Banana → generated clip URL.

Privacy Mode disables all of these at the toggle.

How this looks in practice

In day-to-day use, the easiest way to understand AetherCut’s data collection is to separate “account and billing records” from “editing content.” If you sign in with Google, AetherCut needs enough information to recognize your account later, show basic profile details in the app, and determine whether you have free or Pro access. If you upgrade, it also needs the Stripe identifiers and transaction status required to confirm payment and handle refunds or disputes correctly. That is very different from storing the actual media you work on. AetherCut’s policy here is narrow by design: it keeps the account, payment, and limited product-operation data it needs, while avoiding storage of your raw video content and full IP addresses.

A common misunderstanding is assuming that any AI-assisted action means all project materials are permanently copied into a database. The important distinction is between temporary processing and retained records. For example, generating captions requires audio to be processed so text can be produced, but the retained item is the resulting subtitle text, not a permanent archive of the extracted audio. The same logic applies to other AI features: the system keeps the useful result you asked for, rather than the full source input wherever possible. If you want to avoid that third-party processing entirely, Privacy Mode is the setting to check before using those tools.

Another practical point is that analytics-style page view data is intentionally limited. Seeing that someone visited a path from a referrer host at a certain time helps operate and improve the site, but that record is not tied back to your signed-in account profile. That means browsing data and account identity are kept more separate than many users expect, which matters if you are trying to evaluate how much of your activity is actually being linked together.