Does a transcription site upload your files? How to check
Updated October 31, 2026
Recordings are often private: a doctor's appointment, a job interview, a client call, a lecture you agreed not to share. Most online transcription services work by uploading the file to their servers, which is fine if you trust them and know what they keep. Some tools, including FreeTranscribe, process the file inside your browser instead. Either way, you don't have to take a website's word for it. Here are three checks, from easiest to most thorough, and what we found when we ran them on our own tool.
Why many tools need your file
Speech recognition takes a lot of computing. The traditional setup runs it on the company's servers, so your recording has to travel there first. That isn't automatically bad: server tools can be faster on long files and add features like speaker labels. But it means a copy of your recording exists on someone else's computer, at least for a while, and you're relying on their policies for what happens to it. Tools that run the model in your browser skip the trip, and you can verify that yourself.
Check 1: the offline test (anyone can do this)
If a tool can transcribe your file with the internet switched off, it can't be sending the file anywhere. For a browser-based tool:
- Open the tool and transcribe a short, unimportant clip first. This makes the page download its AI model.
- Leave that tab open and turn on airplane mode, or switch off Wi-Fi and unplug the network cable.
- In the same tab, transcribe the file you actually care about.
- Close the tab before you go back online. That rules out a page saving the file to send later.
If the transcript appears, the work happened on your device. If the tool shows a network error or just spins, it needs a server. On upload-based services this test simply fails, which tells you what you need to know.
Our result: we transcribed a test video, switched the browser offline, and transcribed it again in the same tab. The second run finished normally with the same transcript, word for word.
Check 2: watch the network panel (desktop browsers)
Every browser has developer tools that list each request a page makes. You don't need to be a developer to read the part that matters here.
- Open the tool, then open developer tools: F12 or Ctrl+Shift+I on Windows and Linux, Cmd+Option+I on a Mac.
- Go to the Network tab and tick Preserve log (Chrome and Edge) or Persist logs (Firefox).
- Transcribe a file and watch the list.
- Look for requests whose method is POST or PUT (right-click the column headers to show a Method column). Those are the ones that send data. Click one and open its Payload or Request tab to see what it sends.
An upload stands out: a request to the tool's own servers that takes a while and carries about as much data as your file. Small POST requests of a few hundred bytes are usually analytics or ads, which most free sites use. Those should contain page and event information, not your recording.
What we found on FreeTranscribe
We recorded every request the page made while transcribing a test video in Chrome, from loading the page to finishing the transcript. Here's the summary:
| Stage | Requests to | Data sent in request bodies |
|---|---|---|
| Loading the page | freetranscribe.net (the page itself), plus analytics and ad scripts | 0 bytes |
| First transcription | huggingface.co (8 model files) and cdn.jsdelivr.net (the speech engine), plus ad requests | 0 bytes |
| Second transcription, offline | nothing reachable (ad requests failed) | 0 bytes |
The model and engine requests are downloads to your browser, the same for every visitor and cached after the first time. The site also uses Google Analytics, which records anonymous usage events such as "a transcription started" along with coarse facts like the file type and size, never the file name, audio or transcript. Our privacy policy lists exactly what's collected. If you'd rather send nothing at all, an ad or tracker blocker doesn't stop the tool from working.
Check 3: read what the service says it keeps
For tools that do upload, the privacy policy and terms tell you what happens next. Look for:
- Retention: how long recordings and transcripts are stored, and whether you can delete them yourself.
- Training: whether your files may be used to improve their models, and whether you can opt out.
- Sharing: which other companies (cloud and AI providers) process the audio.
- Location: which country the servers are in, if your work has rules about that.
- An account requirement: if you must sign up, your uploads are tied to your identity.
Vague wording such as "we may use content to improve our services" usually means yes to training. For legal, medical or other confidential recordings, check your organisation's rules before uploading anything.
When on-device transcription is the right choice
- interviews, research and journalism with promises of confidentiality;
- meetings and calls covered by an NDA or internal policy;
- health, legal or personal recordings;
- anything you'd simply rather not hand to a stranger.
The trade-off is that your device does the work, so very long files are faster on a computer than a phone. Our accuracy test shows what quality to expect.
Try FreeTranscribe, and feel free to run all three checks on it.