Trust
How Zıpır processes your documents.
We do not train models on your documents.
Nothing is produced until you approve.
You can delete everything at any time.
A job's approval screen, from the live product (the panel's interface is in Turkish). The scenes are listed with their quotes and source lines; no frame is produced until you approve.
On this page9 sections · 4 min read
How your documents are processed
You upload your document so that Zıpır can make a video from it and check every claim against the source. We use it for nothing else. We do not sell it, publish it or reuse it for anyone else.
TLS is used everywhere in transit. At any time you can download the video (including the Turkish subtitles), the proof report and the video's text.
KVKK
We process personal data in line with Law No. 6698 (the Turkish personal data protection law, KVKK). For personal data in your document, you or your organization is the data controller and Zıpır is the processor acting on your instruction; this split of roles is written in the terms of use. On corporate plans we also sign a data processing agreement (DPA). The privacy notice: KVKK privacy notice. Which data sits where: the privacy page.
Location: The application servers, the database and the file storage sit in data centers in Europe; the companies that run them are based abroad, so this counts as a transfer abroad. The model and voice providers are companies in the US, the email provider is a company in the EU. Hosting in Turkey is an option for corporate customers who ask for it, but it is not ready yet.
Who processes the data
Instead of provider names we write what each group does, which country it is in and what goes to it. Each group processes your data only to serve us; the providers' contract terms do not allow them to use your documents to train models.
- cloud hosting provider (Europe): application servers, database, job queue and server logs; database and file storage service (Europe): videos, audio clips and uploaded documents.
- AI model providers (US): text passages chosen from the document; the providers' terms prohibit training models on customer data. The AI providers' contract terms exclude using submitted content for model training; inputs and outputs are deleted on the provider's side within 30 days at most.
- narration voice provider (US): scene sentences only; the document file, user ID and email address are not sent.
- transactional email provider (EU): the recipient address and email text for the sign-in link, account confirmation and admin alerts.
- payment processor (US; only once purchasing is open): name or company name, email address, billing address, tax ID if any, and card details.
What goes to each of them is written on the privacy page.
Model training
None. We do not train models on your documents, and the AI providers we use work under terms that prohibit training on customer data; our model requests go to their direct API.
Retention and deletion
Your document is yours until you delete it, and you do the deleting; you do not need to write to us. You delete a job from the job's page in the panel: the source file, scene plan, audio, video, proof report and edit frames go together. You delete your whole account from the Account page. What remains after deletion (such as usage records without content) is listed openly on the privacy page.
Access
Private for now: our team makes your first video together with you; only our team sees your documents.
So the team is small, but control of access is yours. You delete a job or your account in the panel yourself; the deletion is entered in the audit log and the files go together. On corporate plans we sign a data processing agreement (DPA). What is deleted and what remains is written line by line on the privacy page.
Reporting a security vulnerability
Write to [email protected]. The answer comes from the person who does the work. Our contact details are also in the security.txt file.
Accuracy
Every claim in a Zıpır video is tied to the source line it came from, and nothing is produced until you approve the scene plan. See an example.
Editorial responsibility
The text on this site is written by the Zıpır team, that is, VitrinAI Digital Studio; the publishing organization is responsible, not one person. Every fact on the pages is tied to a source: the output of the live product, the code itself or an official text. A sentence with no source is not written; topics not yet verified live are marked on the page as "verification in progress".
If you see a mistake, write to [email protected]. The "Last updated" date at the bottom of the corrected page changes. Texts that contain legislation are quotations, not legal opinion.