Free alternatives · Work and productivity
DictationDesk
Hold a hotkey, speak, and clean punctuated text appears wherever your cursor is, in any Windows app.
A free alternative to Wispr Flow, built by Surge Software. Free. You pay only what third-party services cost us, at cost.
Coming soon
The problem
What it solves for a business
For: People who write a lot on Windows (email, chat, documents, code) and teams that want a shared dictionary and snippets.
- Typing long emails and messages is slow, and built-in dictation leaves fillers and false starts in the text.
- Popular dictation apps cost $15 a month and the free tiers run out within the week.
- Some dictation apps use a lot of memory, work only online, or have sent screenshots of the screen to the cloud.
How it works
DictationDesk, step by step
- 01
Hold the hotkey
Hold Ctrl+Win (or any chord you set) in Gmail, Slack, Word, VS Code or Notepad.
- 02
Speak
The tray app records your voice; whisper mode raises quiet speech.
- 03
Transcribe on your PC
The bundled offline whisper.cpp engine turns speech into text with no internet needed.
- 04
Clean up
Fillers and false starts are removed, "no wait" keeps only the correction, spoken punctuation is applied, and the text is formatted for the app you are in.
- 05
Type it in
The text is pasted at your cursor and your clipboard is restored.
Comparison
DictationDesk vs Wispr Flow
| DictationDesk | Wispr Flow | |
|---|---|---|
| Price | Free. You pay only what third-party services cost us, at cost. | $15/month, or $12/month billed yearly (checked 24 Oct 2026) |
| Free tier | Unlimited offline dictation | 2,000 words/week |
| Offline on-device mode | Yes (whisper.cpp bundled) | No, cloud only |
| Idle memory on Windows | 12.7 MB working set (measured) | About 800 MB (community reports, unverified) |
| Screenshots of your screen | Never; context is the app's process name only | Sent screenshots in 2024; now reads nearby text through accessibility APIs |
| Teams | Free, up to 50 members, shared dictionary and snippets | About $10 to $12 per seat, 3-seat minimum (sources conflict) |
| Platforms | Windows only (installer not yet published) | Windows, Mac, iOS, Android |
Wispr Flow details are taken from its public website and may have changed since ($15/month Pro, or $12/month billed yearly ($144/year); free tier 2,000 words/week (checked 24 Oct 2026)).
Why switch
Advantages over Wispr Flow
Works with no internet
Dictation runs on your own PC with whisper.cpp, so it keeps working offline and costs nothing per word.
Light on your laptop
The tray app is built in Rust and used 12.7 MB of memory at idle in our measurement.
Private by default
No audio is stored, no text is kept on the server unless you opt in, and the app never takes screenshots.
Corrections handled
Say "at five, no wait, at six" and only "at 6" is typed. Fillers like "um" and "uh" are removed.
Features
What it does today
- Push-to-talk into any Windows app with Ctrl+Win or a custom chord.
- Offline on-device transcription with whisper.cpp (English model bundled).
- Filler-word removal and corrections ("no wait", "scratch that").
- Automatic and spoken punctuation ("comma", "full stop", "new line").
- Formatting per app: email, chat, code (camelCase, snake_case, symbols) or document, based on the app in front.
- Personal dictionary and snippets, synced from the account site; teams of up to 50 share them.
- Whisper mode for quiet speech, and local history that can be switched off.
- Small footprint: 12.7 MB idle working set measured for the tray app; no audio stored and no screenshots taken.
Use cases
How teams use DictationDesk
Email replies
A manager holds Ctrl+Win in Outlook, speaks a reply, and gets a clean message with a greeting and sign-off.
Chat on the go
A support agent dictates casual Slack or WhatsApp Web replies without trailing full stops.
Coding notes
A developer dictates in VS Code and gets snake_case or camelCase names and spoken symbols.
Team terms
A sales team shares product names in one dictionary so everyone's dictation spells them the same way.
Under the hood
Where AI is used, and where it is not
The model
Speech-to-text runs offline on the user's PC with whisper.cpp (base.en bundled). Optional AI polish and command mode use Qwen on the Surge AI server, with Claude as fallback on a small share of calls.
Rules and code
A deterministic on-device cleaner removes fillers and stutters, applies corrections, spoken punctuation and per-app formatting. The app's context is only the foreground process name.
People in the loop
The user chooses when to dictate by holding the hotkey and sees the text appear at the cursor to edit as usual. AI polish is off by default.
Book a working session
A working session with our architects on one of your use cases. You bring a process and sample documents; we leave you with a scoped plan: model, hosting, integrations, timeline and cost range.