Privacy policy

Version of 13 September 2026. The French version prevails in case of discrepancy.

What we collect, and why

For the subscription: the e-mail address and subscription identifier passed on by Stripe after payment, and the card fingerprint Stripe computes (never the card number), kept so that each card gets only one trial. We never see or store card numbers. Legal basis: performance of the contract.

For the service: the files you entrust (a reduced copy of your video, music, texts), the instructions you write to the studio, and the results produced. Legal basis: performance of the contract.

For metering: minutes of video processed, number of generations and images, per subscription and per month. This is what enforces the volume of your plan.

For running the site: technical server logs (IP address, page requested, date), kept 30 days. Legal basis: our legitimate interest in securing the service. The site uses no advertising cookie and no third-party audience measurement. We count page views ourselves and, on the home page, an anonymous summary of each visit: the name of the site you came from (never the full address), the screen type, the parts of the page shown, the buttons clicked (never what you type), the time spent by range and how far the video was watched. No cookie, nothing stored in your browser and no third-party service: your IP address is never kept, only a daily fingerprint that cannot be traced back, and we only keep daily aggregated counters, which cannot be used to follow a visit.

Where your files go

Your browser makes a reduced copy of your video; the original does not leave your computer. The copy and the working files are stored in a private space (Cloudflare R2, European Union) and processed on servers rented in the European Union and the European Economic Area (Runpod: Romania, Iceland). A reduced copy of your video (picture and sound), images extracted from it and your instructions may be passed to Anthropic and Google for analysis, writing and narration; the text to be spoken is passed to ElevenLabs for the voice. Transcription runs on our rented servers. Under the business terms we have signed with each of them, these services do not keep your content to train their models. Searching for archives and free footage queries Wikimedia Commons and Pexels: those requests carry keywords drawn from your subject, never your files.

If you connect your YouTube channel

Connecting your channel is optional: the studio works without it. If you do, you allow Kutavo to do three things, and nothing else:

• Upload the videos you ask to publish, and their thumbnail (Google permission youtube.upload). Kutavo never publishes before you have approved the title, the description and the visibility.
• Read your channel information: its name, subscriber and view counts, and your 50 most recent videos with their title, date, duration and public counters (views, likes, number of comments) (youtube.readonly).
• Read your YouTube Analytics statistics: views and minutes watched per day, average view duration and percentage, subscribers gained and lost, shares, and traffic sources (yt-analytics.readonly).

This data is used only for what you see in the studio: publishing your videos, showing your channel dashboard and giving you advice drawn from your numbers. It is read when you connect your channel and when you ask for a refresh. Kutavo never edits or deletes an existing video, never reads the content of any comment and never posts on your behalf.

What is shared. To write the three-sentence advice about your channel, a summary of these numbers (channel name, counters, titles of your recent videos) is sent to our AI provider (Anthropic, or Google Gemini as a fallback), which does not keep it to train its models. Your access token is never passed on. Nothing is sold, used for advertising, used to train a model, or transferred to anyone else. Kutavo's use and transfer of information received from Google APIs will adhere to the Google API Services User Data Policy, including the Limited Use requirements.

How long, and how to stop it. The numbers read are kept for the duration of your subscription, to show how your channel evolves. The Disconnect button in the studio immediately revokes the access at Google and deletes your token and statistics. When the subscription ends, the access is revoked and everything is deleted within 30 days. You can also remove the access at any time, without going through us, from your Google account permissions; the studio will notice and tell you. By connecting your channel, you use the YouTube API Services: the YouTube Terms of Service and the Google Privacy Policy then also apply.

How we protect your data

• Encryption in transit: every connection (your browser, the studio, our exchanges with Google, storage and AI providers) goes through HTTPS (TLS).
• Encryption at rest: files in the storage space (Cloudflare R2) are encrypted at rest.
• Your YouTube access, a sensitive piece of data: the token Google gives us is kept on our server in France (OVH), in your own space, in a file readable only by the service account. It is never displayed, never included in backups, never sent to an AI provider or any third party, and is only used for the Google API calls you trigger.
• Isolation: each subscriber has their own space, and every request is checked so that only your files are served. Sign-in links sent by e-mail work only once. Payments are handled by Stripe: no card number goes through us.
• Server access: restricted to the publisher, with SSH keys only (password login disabled), behind a firewall that opens only the web and administration ports. Service keys are kept out of the code.
• Minimisation: we request only the permissions needed for the features above, and working files are deleted at the latest 30 days after delivery.
• Incidents: in the event of a data breach, we notify the French data protection authority (CNIL) within 72 hours, and the people concerned when the risk is high, as the GDPR requires.

For how long

Working files and results are deleted at the latest 30 days after delivery, giving you time to collect them. Your conversation with the studio, your ongoing edits and, if you created one, your cloned voice are kept for the duration of the subscription, then deleted within 30 days of its end. A cloned voice is personal data: you only clone your own voice, or that of a person who gave you consent.

Your rights

You can access your data, correct it, have it deleted, ask for a copy or object to a processing, by writing from the contact page. We answer within a month. You can also lodge a complaint with the French data protection authority, the CNIL (cnil.fr).

Data controller

Dylan Becerra, sole trader, 64170 Lacq, France, SIREN 952 822 633. Processors: OVH (hosting, France), Cloudflare, Inc. (storage, EU servers) and Runpod, Inc. (computing, EU servers), US companies covered by standard contractual clauses, Stripe (payment, Ireland), Google, Anthropic and ElevenLabs (AI, United States, standard contractual clauses), Wikimedia Foundation and Pexels (search for archives and free footage, United States: keywords only, no file and no personal data).