Help & FAQ

No. PrivateTranscribe runs on your processor, so any modern Windows PC works fine.

If you have an NVIDIA graphics card (GTX 10-series or newer), the app quietly downloads a GPU engine. Combined with local processing, that means no upload round trip before transcription starts.

AMD and Intel graphics cards run on the processor for now. GPU support for those is planned. With a local Whisper model selected, raw audio stays on your PC during transcription; cloud transcription is an optional choice.

No for the core local dictation flow. Once installed, local transcription runs on your machine with local Whisper models.

Internet is used only for explicit online features such as checking for app updates, optional Pro license activation, and any cloud provider you choose to enable for transcription or AI enhancement.

The installer is signed by Julsgaard Products, so Windows shows a real publisher name instead of an unknown one.

SmartScreen reputation is separate from signing. It builds as more people download each release, so a new version can still trigger a warning. Click "More info" then "Run anyway" to proceed.

This happens occasionally, and it is a false positive. PrivateTranscribe does two things that look like malware to a scanner watching behaviour alone. It listens for your dictation hotkey system-wide, which is the same mechanism a keylogger uses, and it types the finished text into whatever app you had open, which is the same mechanism input injection uses. Both are the product doing its job.

If the installer or the app gets quarantined, restore it and add PrivateTranscribe to your exclusions. In Bitdefender that lives under Protection → Antivirus → Settings → Manage exceptions. Most other scanners have an equivalent under a name like exclusions, allowed apps, or trusted files.

We report every Windows release to Microsoft and Bitdefender as a false positive, and detections usually clear within a few days. Scanners judge each file by its hash, so a new release can trip the same warning again before it has been reported.

If your scanner flags something and you want it looked at, send the detection name and the version you installed to support@privatetranscribe.com and we will get it submitted.

If you have a modern CPU (last ~5 years): Start with small or large-v3-turbo.

If you have an NVIDIA GPU: Go straight to large-v3-turbo. CUDA is included in Starter because speed and quality are the proof.

If your machine is older: Start with tiny or base and work up.

PrivateTranscribe auto-detects your hardware and recommends the best model for you.

App data is stored locally in %APPDATA%\PrivateTranscribe. This includes your transcription history, settings, dictionary, and correction memory. PrivateTranscribe does not sync that database to the cloud.

Yes. PrivateTranscribe works in any text field, including VS Code, Cursor, JetBrains IDEs, and terminals. The global hotkey works system-wide.

Pro is planned to include Smart Context, which can use your active app/window context to improve transcription hints without browser or IDE extensions.

Starter includes 1,000 transcribed words per day, every Whisper model, CUDA acceleration, and Read Aloud. Pro is a one-time purchase that removes the daily dictation limit and includes Converse.

Dictation enhancement, Correction Memory, Smart Context, and Action Engine remain available only to approved testers until stable. Coding prompt shortcuts are available without tester access. Buying Pro does not grant tester access.

Go to privatetranscribe.com/license, enter the email you bought with, and we will send the key again. It only ever goes to that address.

If nothing arrives, check spam, then try the other address you might have used. Failing that, email support@privatetranscribe.com.

30-day money-back guarantee, no questions asked. Email support@privatetranscribe.com and we'll process your refund within 5 business days.

Yes. Whisper auto-detects 99 languages. Set your language in Settings → Speech, or leave it on "Auto-detect". You can also enable "Translate to English" to speak in any language and get English output.

Switch to a smaller model. If you're using large-v3, try large-v3-turbo or small.

Close heavy apps. Whisper uses your CPU/GPU - other heavy workloads compete for resources.

Enable CUDA. If you have an NVIDIA GPU, make sure CUDA is being used (check Settings → Models).