Privacy & Security
How PrivaSpeech handles dictation
A plain-language explanation of what happens to your dictation audio, where transcription happens, and what still requires internet. No legalese, and no hand-waving.
The core principle: dictation runs on your Windows machine
When you dictate with PrivaSpeech, transcription happens on your Windows PC. Dictation audio is not sent to third-party cloud transcription APIs.
0
Audio sent to cloud for transcription
0
Third-party transcription APIs processing your voice
Local
Dictation processing happens on your PC
Step-by-step: what happens during live dictation
You press the hotkey
PrivaSpeech starts recording from your microphone. The audio is captured locally on your Windows machine and stored temporarily during transcription.
You release the hotkey
Recording stops. The audio file is passed to the local Parakeet V3 transcription engine running on your Windows machine.
Transcription happens on your PC
The ONNX-based Parakeet V3 model processes the audio on your hardware. Depending on your system, this uses your CPU or local acceleration on your machine rather than external cloud compute.
Text is copied to clipboard
The transcribed text is placed in your Windows clipboard. You can paste it into any application.
Temporary audio is deleted
The temporary audio file created during recording is automatically deleted after transcription completes.
Data flow diagram
Input
Processing
(Parakeet V3 + ONNX)
Output
What this means in practice
Good fit
PrivaSpeech is a strong fit when you want faster drafting on Windows without routing dictation audio through external transcription services. That often matters for client emails, matter notes, draft letters, and browser-based case systems.
Important caveat
On-device transcription can reduce exposure compared with cloud-first tools, but it does not replace your own security, confidentiality, or compliance review. Evaluate the workflow against your firm's requirements before relying on it for production legal work.
What does need internet?
Dictation itself runs locally after setup, but a few parts of the product still require internet connectivity. Here's the complete picture:
- •Initial model download (~465MB, happens once during setup)
- •Application updates (optional, when you choose to install them)
- •License activation and validation
Works offline after setup and activation
- •Voice recording
- •Speech-to-text transcription
- •Clipboard integration
- •Hotkey functionality
PrivaSpeech vs. cloud dictation
Here's how PrivaSpeech compares to typical cloud-based speech-to-text services:
| Feature | PrivaSpeech | Cloud dictation |
|---|---|---|
| Audio processing location | Your PC | Remote servers |
| Works offline | Yes (after setup and activation) | No |
| Audio sent to third parties | No | Yes |
| Latency | Low (local) | Variable (network-dependent) |
| Data retention by provider | None | Varies by provider |
| Pricing model | $29/moor $249/yr | Often per-minute or per-word |
| BAA/compliance certifications | Not currently offered | Some providers offer |
What we don't do
We don't stream your audio to external transcription APIs
We don't store recordings after transcription completes
We don't use your voice data to train models
We don't sell or share your transcription data
Common questions
Can I verify that no data is being sent?
Yes. You can use network monitoring tools like Wireshark or Windows Resource Monitor to verify that no audio data is transmitted during transcription. After the initial model download, you can even disconnect from the internet entirely.
What data do you collect?
We collect license validation data to verify your subscription status. Audio and transcription content are never transmitted. See our privacy policy for details.
Is this suitable for HIPAA/SOC2/other compliance?
Local processing reduces third-party data exposure, which may support certain compliance requirements. However, we do not currently hold formal certifications. Consult your compliance team for your specific use case.
How big is the model download?
The Parakeet V3 model is approximately 465MB. It's downloaded once during initial setup and stored locally for offline use.
Try it yourself
Start with a free 30-day trial. Experience local dictation on your Windows PC.