01
100% local transcription
Whisper runs on your CPU. No audio ever leaves your machine, no API keys, no rate limits. Works on a plane.
Now in early access
Speak. It types.
SpeakCLI is a push-to-talk dictation tool for Windows and Linux. Press Ctrl+. speak, and your words land wherever your cursor is — Claude Code, Warp, your editor, anywhere. Whisper runs on your machine. No audio ever leaves your device.
- ✓ 100% local transcription
- ✓ 99 languages, free model included
- ✓ No subscription
- ✓ Pay once, own it
⟶ Always-on overlay
deploy the staging environment
02 / Features
/specs
Built for the way developers actually work.
Six things that matter. Six things we ship. No empty pricing tiers, no "Enterprise" upsell calls. Everything below is in the box.
02
Voice macros
Map spoken phrases to text. "Deploy staging" → npm run deploy:staging. Exact match or regex.
03
CLI-aware dictionary
Tell SpeakCLI about kubectl, terraform, npx. Whisper picks them up cleanly instead of guessing.
04
Code mode
Lowercase, strip spoken punctuation, ready to paste into a terminal. Toggle with a second hotkey.
05
End-to-end encrypted sync
Macros, dictionary, theme — synced across your devices via Argon2id + AES-256. We can't read it.
06
Multilingual — 99 languages
Whisper supports ~99 languages out of the box. The multilingual base model is free alongside the English-only one. Speak any language and get text back in that language — or flip on Translate mode and get English text instead.
03 / Workflow
/howto
Three steps. Two seconds.
From hotkey to typed text. Real-time, no cloud round-trip, no model loading delay (we keep Whisper warm in the background).
01
$ press Ctrl+.
Press the hotkey
Press Ctrl+. anywhere in Windows or Linux. The floating overlay turns on and shows you it's listening.
02
$ speak naturally
Speak
Talk like you would to a colleague. Pauses are fine. Whisper handles tech vocabulary, accents, and incomplete sentences.
03
$ release Ctrl+.
Words appear
Press the hotkey again. The text lands at your cursor — terminal, IDE, browser, anywhere with a text input.
04 / Comparison
/vs
Why not the others.
Push-to-talk, custom vocab, a hundred languages — every product in this category does the basics. The real differences are who runs the model, who keeps your data, and how often you pay for the same software.
Capability
SpeakCLI
⟶ this product
Wispr Flow
SuperWhisper
Aqua Voice
Push-to-talk hotkey dictation
✓
✓
✓
✓
Multilingual (50+ languages)
~99
100+
100+
~50
Custom vocabulary / dictionary
✓
✓
✓
✓
Where we differ
Runs fully offline — audio never leaves the device
✓
—
◐
—
Works on Linux
✓
—
—
—
No account or sign-in required
✓
—
—
—
Free tier — no word or time cap
✓
2k words/wk
◐
1k lifetime
Pro is one-time, not a subscription
$9.99 once
—
$249 once
—
Voice macros with regex
✓
—
—
—
Terminal-ready code mode
✓
—
◐
—
◐ = partial / with caveats — see each product's docs.
Based on each product's public information in May 2026. Spot something out of date? Email support@speakcli.dev and we'll update the table.
05 / Pricing
/billing
Free forever. Pro is one-time.
Pay once, own it. No subscription, no usage cap, no per-seat ladder. Founder pricing for the first 100 customers — then we move to standard pricing.
Free
$0 /forever
Everything you need to dictate. No nag screens, no time limits, no asterisks.
- Whisper transcription — English or any of ~99 languages, optionally translated to English
- Single global hotkey
- Auto-stop on silence
- Always-visible overlay
- Windows and Linux
Founder · first 100
Pro
$9.99 one-time · lifetime
Then $19.99 after the first 100 customers.
Unlocks the productivity layer. Buy it once, run it forever, install on up to 2 devices.
- Everything in Free, plus:
- Voice macros (exact + regex)
- CLI-aware dictionary
- Smart post-processing + code mode
- Multi-hotkey profiles
- Custom avatar in the overlay
- Hidden / on-demand / background overlay
- Larger Whisper models for higher accuracy (English + multilingual)
- End-to-end encrypted cross-device sync
- Auto-start with the OS
- Up to 2 devices · Lifetime updates
Lifetime updates · Secure checkout via Polar
06 / Who
/whoami
Built by one developer.
No marketing team, no "we" pretending to be a company. You're buying software from a person. Here's who.
$ whoami crvx14 $ cat ./bio.txt Five years writing TypeScript and Java for a living. Solo on SpeakCLI. $ cat ./why.txt Every dictation tool I tried either shipped my audio to a server I don't control, charged me monthly rent for software running on my own CPU, or didn't exist on Linux. So I built one that does none of those things.
If you find a bug, hit a wall, or want a feature — email me directly. Replies are same-day in practice.
07 / FAQ
/answers
Things people tend to ask.
If something below doesn't cover it, email support@speakcli.dev and we'll get back same-day.
Does my audio go to a server?
No. SpeakCLI bundles a whisper.cpp build and runs it locally on your CPU. Audio never leaves your device. Your transcribed text doesn't leave either — it's typed into the foreground window directly via the OS keyboard API.
Windows says 'Unknown publisher' / SmartScreen blocked the installer. Is this safe?
Yes — this is expected for v1. Click 'More info' → 'Run anyway' to install. The warning appears because the installer isn't code-signed yet (a Windows code-signing certificate costs $200-700/year and we're keeping the founder price low). The binary is the same one signed by our updater key — check `latest.json` for the signature. As more people install, SmartScreen learns the binary is safe and the warning disappears on its own. We'll add code-signing once revenue justifies it.
My antivirus flagged the installer. Is it malware?
No. SpeakCLI uses a few OS APIs that overlap with what malware also uses: microphone access, a global hotkey listener, and simulated keyboard input (to type the transcript into the foreground window). Combined with an unsigned binary, a handful of niche engines flag this heuristically. Verify for yourself on VirusTotal — every major consumer AV (Defender, Bitdefender, Kaspersky, Norton, Avast, AVG, McAfee, ESET, Sophos, Malwarebytes) reports the installer clean. If your AV blocked the install entirely, whitelist the SpeakCLI install folder and we'll re-scan on the next release.
Verify on VirusTotalWhy not a subscription?
Because every developer tool we love (Sublime, JetBrains in the old days, Alfred, Keyboard Maestro) was a one-time purchase. Subscriptions buy you cloud sync and recurring features; if the product is local-first, the subscription is just rent. We'd rather charge a fair one-time price and earn future revenue from upgrades.
What's the device limit?
A Pro license activates on up to 2 devices simultaneously. You can deactivate from any device (Settings → Account → License → Manage Devices) to free a slot. Free downloads are unlimited — no account or activation required.
Can I get a refund?
Yes — within 14 days, no questions asked. Email support@speakcli.dev with your Polar order ID and we'll refund through Polar within one business day. Money lands back on the card you paid with.
When will macOS be supported?
macOS is on the roadmap but not committed to a date. It requires an Apple Developer account ($99/yr) and the build pipeline + notarization workflow needs to be set up. The codebase is mostly portable — most macOS-specific work is around audio device permissions and the menu bar integration.