Voice dictation for Windows 10/11
SparkFlow puts clean text at your cursor in whatever Windows app has focus. The speech models run on your machine, so your audio never leaves it — and it works with the Wi-Fi off.
7 days, fully featured · no card, no account · then $5/month
What you said
Fillers dropped, pasted at the caret
How it works
Press your hotkey in any app — editor, browser, Slack, a terminal. Push-to-talk, or switch it to toggle mode.
Whisper on your GPU, Parakeet on CPU. Dozens of languages. Everything stays inside your machine, online or off.
Release the key and the cleaned text lands at your cursor — pasted the way that app expects, with your clipboard left untouched.
Private by architecture
There is no upload step to trust. The speech models are files on your disk; transcription happens in a process on your PC, and the result goes straight to your cursor. No account to create, nothing to sign in to.
If you want to check: turn off your Wi-Fi and keep dictating. It works exactly the same.
| SparkFlow | Cloud dictation | |
|---|---|---|
| Where audio is processed | On your machine | On their servers |
| What you pay | $5/month, flat | Often $15/month and up |
| Without internet | Works | Doesn't work |
| Account required | No | Usually |
Cloud tools are often faster to set up and need no local GPU. This is the trade we made, not a verdict on theirs.
What it fixes on the way out
Filler words
so um basically we should ship friday
We should ship Friday.
Stutters
the the the invoice is is wrong
The invoice is wrong.
Custom vocabulary
deploy it to coobernetties
Deploy it to Kubernetes.
Your own list of names and jargon, fuzzy-matched.
Per-app paste
SparkFlow pastes the way each app expects, and falls back to a second method automatically when an app blocks the standard one. Whatever you had copied stays on your clipboard.
History
Every transcription is kept locally, so you can search what you said last Tuesday and copy it again. It's a file on your disk — delete it whenever you like.
Pricing
One plan, one price. No upcharges, no per-seat tiers, no hidden fees — and nothing is feature-gated during the trial, because the trial is the full product.
Cloud dictation tools typically run $12–15/month. SparkFlow is $5 — and your audio never leaves your machine.
Free trial
7 days
Fully featured. No card, no account.
Subscription
$5/month
Billed monthly. Cancel any time.
Questions
Whisper models run GPU-accelerated when you have a GPU; Parakeet handles CPU-only machines. Larger models are more accurate and slower — you choose which one to load, and the custom vocabulary catches the names and jargon a general model tends to miss.
Dozens, through the Whisper models — the model you load determines the languages available to you.
Windows 10 or 11, x64. A GPU is optional and makes transcription faster; without one, the CPU model does the work.
Yes. The models run locally and transcription happens on your machine, so there is nothing to send. Disconnect from the internet and dictation still works — that is the simplest way to verify it yourself.
During the beta the installer isn't code-signed yet, so SmartScreen shows a blue "Windows protected your PC" screen. Click More info, then Run anyway. It's a reputation warning, not a detection — and it will go away once signing is in place.
Refunds: [REFUND-POLICY]. You can cancel any time and keep using SparkFlow until the month you paid for is up — and the 7-day trial is there so you can be sure before paying anything.
7 days · no card · no account