QuietSteno

Home / Guides / Is online transcription private?

Is online transcription private?

It depends on where the speech recognition runs. If your recording is uploaded, its privacy rests on the provider’s policies. If it is processed on your device, there is no copy to protect.

Most online transcription services are not private in the strict sense: your recording is uploaded to the provider’s servers, processed there and usually stored for some period. How long it is kept, who can access it and whether it is used to train models depends on that provider’s terms. On-device transcription tools avoid this by running the speech model in your browser, so the audio never leaves your computer.

What happens when you upload a recording

A typical cloud transcription service works like this: the file travels over an encrypted connection to the provider’s servers, a speech model processes it, and the transcript and usually the audio are saved to your account so you can come back to them. Encryption in transit protects the file on its way there. It does not change the fact that the provider now holds a readable copy.

From that point, privacy is a matter of policy and security practice. Reputable providers take both seriously, but you are relying on their choices, not on a technical guarantee.

Questions to ask a cloud provider

  • Retention: how long are audio and transcripts kept, and does deleting them remove backups?
  • Training: are customer recordings used to improve models, and is that opt-in or opt-out?
  • Human review: can staff or contractors listen to recordings for quality control?
  • Subprocessors: which other companies handle the data, and in which countries?
  • Agreements: if you handle personal or health data, will the provider sign the data processing or business associate agreement your regulations require?

The answers are in the privacy policy and terms of service. They differ a great deal between providers and between free and paid plans.

How on-device transcription differs

Speech models have become compact enough to run inside a web browser. An on-device tool sends the model to you instead of sending your audio to the model. The page downloads the model files once; after that, your browser reads the recording from disk and does all the processing with your own processor or graphics chip.

No audio is transmitted, so there is nothing for a provider to retain, review or train on. This is how QuietSteno works.

How to check a tool’s claim yourself

You do not have to take “runs locally” on trust:

  1. Watch the network. Open your browser’s developer tools (F12), go to the Network tab and start a transcription. A local tool shows model files being downloaded and no large outgoing request with your audio.
  2. Disconnect. Once the model has loaded, turn off Wi-Fi and transcribe a file. A cloud tool stops working; a local one continues.
  3. Look at the speed of the start. A cloud tool has to upload first, which takes time on a large file. A local tool begins as soon as the file is read.

The trade-offs

Cloud serviceOn-device tool
Where audio is processedProvider’s serversYour device
Copy held by a third partyYesNo
AccuracyHighest, using very large modelsGood for clear speech; more errors on hard audio
SpeedConsistentDepends on your hardware
Extras such as speaker labels and summariesCommonLimited
CostFree tier, then paidFree

When each makes sense

For a public podcast or a marketing video, privacy is not a concern and either approach is fine. For interviews with sources, patient or client conversations, internal meetings, legal matters and research covered by an ethics protocol, keeping the recording on your own machine removes a whole category of risk and paperwork.

Local processing does not make everything else safe. Your computer still needs a password and disk encryption, and transcripts you paste into other online tools are subject to those tools’ policies.

Transcribe a file now

Questions and answers

Is it safe to upload confidential audio to a transcription website?

It can be, if the provider’s retention, access and training terms meet your obligations, and you have checked them. Where you cannot accept any third-party copy, use a tool that processes the audio on your device.

Do transcription services use my recordings to train AI?

Some do, some do not, and some allow you to opt out. It is stated in each provider’s terms and often differs between free and paid plans.

How can I transcribe audio without uploading it?

Use an on-device tool. In a browser-based one such as QuietSteno, the speech model is downloaded to your browser and the recording is processed locally.

Is on-device transcription compliant with GDPR or HIPAA?

Local processing avoids sharing the recording with a processor, which simplifies compliance. It does not replace your own obligations, such as having a lawful basis, securing your device and following your organisation’s policies.

Related

Last updated 13 November 2026.