Frequently Asked Questions

What is PrivaSpeech?

PrivaSpeech is a Windows desktop application that transcribes speech to text using local AI processing (powered by Parakeet v3 technology). Your dictation audio is processed on your device. It is a strong fit for small law firms and other privacy-sensitive professionals who want faster drafting without routing audio through cloud transcription services. Use a global hotkey for live dictation or import supported audio files for local transcription.

How does PrivaSpeech work?

It's simple:

1. Press Ctrl+Space anywhere on your computer to start recording
2. Speak into your microphone
3. Press Ctrl+Space again to stop recording
4. The transcribed text is automatically copied to your clipboard - just paste it into Word, Outlook, your browser, or anywhere else you work

After the initial setup, dictation works locally on your Windows machine. That makes PrivaSpeech useful for confidentiality-sensitive work and for environments where internet access is limited.

Who is PrivaSpeech a good fit for?

PrivaSpeech is the best fit for Windows users who want a clipboard-first dictation workflow and care about keeping transcription on-device. Today that especially includes small law firms, solo attorneys, paralegals, legal assistants, and other professionals with confidentiality-sensitive drafting workflows.

What makes PrivaSpeech different from other dictation software?

Do I need a $2,000 GPU to run PrivaSpeech?

No. An NVIDIA GPU can help with performance, but PrivaSpeech is optimized to run efficiently on standard modern CPUs with the ONNX runtime. PrivaSpeech works great on Intel i5/Ryzen 5 processors and up. A dedicated GPU is not required.

What are the system requirements?

• OS: Windows 10/11 (64-bit)
• CPU: Modern Intel i5/Ryzen 5 or better recommended
• RAM: 4GB minimum, 8GB recommended
• Storage: ~500MB for app + models (~465MB downloaded on first launch)
• Microphone: Any Windows-compatible microphone
• Internet: Required only for initial model download, license activation, and optional updates

How accurate is the transcription?

PrivaSpeech is powered by Parakeet v3 technology, which delivers strong accuracy for clear microphone audio. Accuracy varies with noise, accents, domain vocabulary, and mic quality.

How does the model download work?

On first launch, PrivaSpeech automatically downloads the required AI models (~465MB) from Cloudflare R2. The download includes:

• Parakeet v3 transcription model
• Silero Voice Activity Detection (VAD) model

All downloads are verified using SHA256 checksums to ensure integrity. Once downloaded, models are stored locally on your device and used for on-device transcription.

Can I customize the hotkey?

Yes! The default hotkey is Ctrl+Space. If that conflicts with another app, PrivaSpeech will try a fallback (Ctrl+Shift+Space). You can also change it in the app settings. The hotkey works system-wide, so you can trigger recording from any application.

Can I import existing audio files?

Yes. PrivaSpeech can transcribe local `.wav`, `.mp3`, `.m4a`, and `.aac` files from the Import audio tab. The file stays on your device during decoding and transcription. Olympus `.dss` and `.ds2` files are not currently supported.

Is my voice data sent to the cloud?

No—dictation audio and transcripts are processed locally.

PrivaSpeech does not upload your dictation audio or transcripts for transcription. After the initial model download, you can use transcription offline. This can be a good fit for:

• Small law firms handling client-sensitive drafting and matter notes
• Privacy-conscious professionals working with sensitive data
• Teams working in locked-down or privacy-sensitive environments
• Anyone who values local processing and data control

Does PrivaSpeech collect any telemetry or analytics?

The desktop app does not collect usage analytics by default. We don't sell your data, and we don't upload your dictation audio or transcripts for transcription.

Can I use PrivaSpeech for confidential legal work?

PrivaSpeech keeps transcription on your device instead of routing dictation audio through external transcription services. That can make it easier to evaluate for confidentiality-sensitive legal workflows, but you should still review it against your firm's own policies, obligations, and risk posture before relying on it for production legal work.

Where is my audio data stored?

Audio is temporarily recorded to a WAV file in your system's temp directory during transcription. This file is automatically deleted immediately after transcription completes. Imported files are read from the location you choose and transcribed locally; current supported import formats do not require a separate temp conversion artifact. No audio is permanently stored by the app. Microphone transcripts go directly to your clipboard, while imported transcripts do not auto-copy by default.

What does private mode do for imported files?

Private mode keeps that imported-file transcript out of local history. The transcription still runs on-device, but PrivaSpeech skips the history save for that import so you can review or copy the text without leaving a local history entry.

How much does PrivaSpeech cost?

PrivaSpeech offers flexible pricing to fit your needs:

• Monthly subscription: $29/month with a 30-day free trial
• Yearly subscription: $249/year with a 30-day free trial

Both options include all features and come with a 30-day money-back guarantee.

What if I'm not satisfied?

We offer a 30-day money-back guarantee. If PrivaSpeech doesn't meet your needs for any reason, contact us for a full refund - no questions asked.

Do I need to create an account?

No account creation required. No login. No email signup. Just download, install, and start dictating. We only need your payment information to process checkout.

Can a small firm evaluate PrivaSpeech for multiple people?

Yes. The simplest path is to start with one user and confirm the workflow in real work first. If your firm wants a guided small-team evaluation or manual billing, contact us and we can help with the rollout path that fits your situation.

How do I install PrivaSpeech?

1. Download the MSI installer (~15MB)
2. Run the installer (standard Windows setup)
3. Launch PrivaSpeech - models will download automatically on first launch (~465MB) and stay on your device forever
4. Press Ctrl+Space to start dictating!

Why do I see a Windows SmartScreen warning?

When installing PrivaSpeech, you may see a Windows SmartScreen popup saying "Windows protected your PC" or "Unknown Publisher". This is completely normal for new software.

Why this happens: Windows SmartScreen warns about any software that isn't signed with an expensive code signing certificate (~$300-700/year). As an indie project, we haven't yet invested in code signing. This warning does not mean the software is dangerous; it means Windows does not recognize the publisher yet.

How to proceed:

1. Click "More info" on the SmartScreen popup
2. Click "Run anyway"
3. The installer will proceed normally

Is it safe? All dictation processing happens locally on your device rather than being sent to external transcription services. You can verify the offline dictation behavior yourself after setup is complete.

Do I need an internet connection?

You only need internet for the initial model download (~465MB), license activation, and optional updates. After setup, transcription runs locally on your device rather than being sent to external transcription services.

What languages does PrivaSpeech support?

Currently, PrivaSpeech supports English transcription. The Parakeet v3 technology supports multiple languages, and we're working on adding language selection in future updates.

Can I use PrivaSpeech with any application?

Yes! Since PrivaSpeech uses a global hotkey and clipboard integration, it works with any application that accepts text input. Use it in Word, Outlook, VS Code, Slack, Notepad, email clients, EMR systems, legal case management software, or any other text editor. The transcribed text is copied to your local clipboard, ready to paste anywhere.

The hotkey isn't working

Try these steps:

• Check if another application is using the same hotkey
• Restart PrivaSpeech
• Try changing the hotkey in settings to a different combination
• Some hotkeys require administrator privileges on Windows

Recording isn't starting

Check the following:

• Verify microphone permissions in Windows Settings
• Ensure your default input device is set correctly
• Use the "Mic Check" feature in the app to test your microphone
• Check that no other application is exclusively using your microphone

Model download failed

If the model download fails:

• Check your internet connection
• Ensure you have enough disk space (~500MB free)
• Try restarting the app - it will retry the download
• Check Windows Firewall isn't blocking the connection
• Contact support if the issue persists

Transcription accuracy seems low

To improve accuracy:

• Speak clearly and at a normal pace
• Reduce background noise
• Use a quality microphone (USB microphones often work better than built-in ones)
• Ensure your microphone is positioned correctly
• Wait for the model to "warm up" on first use (shown in the UI)

Need more help?

If you have questions not covered here, please contact us at support@privaspeech.com. We're here to help!