When you upload a recording to an online transcription tool, your audio and the transcript it produces are personal data, and you deserve a clear answer about what happens to them. This guide explains how RealtimeVoiceKIT handles that data: where it is stored, how long it stays, who can reach it, which outside providers are involved, and what you can do about all of it. Privacy here is built in by design, not bolted on afterward.
The first thing to know is that your files and transcripts remain yours. Uploading a recording does not hand it over to us to do as we please. You can delete an individual transcript at any time, delete your whole account, or export a full copy of your data to take it elsewhere. Anonymous demo transcripts created on the marketing site, where no account is involved, are deleted automatically after a short window, so a quick test does not leave a lasting trace.
Your data is protected both while it moves and while it sits still. Connections use TLS, so audio and text are encrypted in transit between your browser or app and our service. At rest, data is encrypted on Google Cloud, our hosting provider. Internal access to production data is deliberately limited to the people and processes needed to operate the service, not browsed casually by staff.
Like most modern software, RealtimeVoiceKIT relies on a small set of sub-processors to do its job: hosting, payments, email delivery, AI translation, and analytics. Each one operates under a data-protection agreement, and the full current list lives on our Sub-processors page so you always know who is in the loop. The component that converts your speech to text is our speech-to-text processing provider, also covered by an agreement, and the AI behind translation and summaries is powered by leading frontier AI from OpenAI (ChatGPT), Anthropic (Claude), and Google (Gemini).
Two commitments matter most to privacy-conscious users. We do not train models on your audio, video, or transcripts, and we do not sell personal information. Your recordings are processed to give you a transcript and the features you asked for, not quietly repurposed as training material or sold to anyone. Retention follows the same principle: your content stays until you delete it, rather than being kept indefinitely for reasons you cannot see.
Analytics and error monitoring are handled with consent at the center. Those tools receive personal data only after you give analytics consent, and we honor Global Privacy Control (GPC), so a browser-level opt-out signal is respected automatically. This approach is shaped by privacy frameworks such as GDPR and CCPA, which give you rights to access, export, and delete your data, the same rights the product surfaces directly in your account.
If you connect through the developer API instead of the web app, the same handling applies. Your rtvk_ key authenticates your requests, transcripts come back with timestamps and confidence, subtitles export to SRT or VTT, and the encryption, retention, and no-training commitments carry over unchanged. When payments are involved, card data is handled by Stripe, a PCI-DSS Level 1 payment provider, so we never store raw card numbers ourselves.
The short version: your transcription data is encrypted, kept only as long as you want it, never sold, never used to train models, and always under your control to export or delete. If you would rather evaluate this yourself, start on the free plan, run a recording through, then export it and delete it to see the controls in action.
The RealtimeVoiceKIT team writes about audio, AI, and the workflows that turn recordings into reach for the RealtimeVoiceKIT team.