NEWVersion 0.9.6:See your text as you speak

Hold the key. Speak. Let go.

FlowScribe types what you say into any Windows app. Locally on your PC, no account.

Try the free beta See how it works

Windows shows a warning on first launch. Here’s what to do ›

SpokenExample

Hold Ctrl + Win, speak, let go

Speed

You speak faster than you type.

The same text, twice in real time: typed once, dictated once with FlowScribe.

Typed

43words per minute

Dictated with FlowScribe

152words per minute

Dictated in 12.9 seconds, typed in 36.3 seconds.

How this was measured

Dictation: measured in my own FlowScribe history on my PC. 43 dictations in German with 5,827 words between 26 October and 10 November 2026, 152 words per minute on average. The wait until the text is inserted (2.3 seconds on average) is included. Typing: the average from an Aalto University study with 168,000 participants (52 words of 5 characters each per minute), converted to the word length of my dictations. The text is an example, your speed may differ.

Everywhere you type.

No plug-in, no extension. FlowScribe inserts the text wherever your cursor is.

How FlowScribe works

01 · Modes

Writes like you. Differently in every app.

Optimized adds punctuation, removes filler words and smooths the sentence. Original keeps exactly what you said. FlowScribe recognizes the app and picks the mode itself. You can adjust every rule.

02 · Commands

Select text. Say what should happen.

Select text, hold Ctrl Win Shift and say “make this friendlier”. Or shorter. Or in German.

03 · Dictionary

Knows your names and technical terms.

Terms, replacements and snippets. A spoken keyword inserts your address or signature.

04 · Audio import

Voice message in, text out.

WhatsApp voice messages, audio files and even MP4 videos become text, with paragraphs and locally as well. In the clip: a German voice message.

05 · History and stats

See how much typing you save.

Every dictation is searchable, with words per day and time saved. You decide how long entries are kept.

FlowScribe settings page “Modes & Apps”: rules for which mode applies in which app FlowScribe history with the command “Make this friendlier”, selected text and result side by side

Real recordings from FlowScribe 0.9

Private

Your voice stays on your PC.

Speech recognition and text editing run on your computer. FlowScribe only stores recordings if you turn that on yourself.

Local. Even offline.

There is no server your audio goes to. The models live on your computer.

Hold. Speak. Let go.

One shortcut, in every program. A quick tap switches to hands-free.

No account.

Install and get going. No sign-up.

German and English.

Built for German, understands English too, even mixed in one sentence.

Your CPU is enough.

Runs smoothly without a graphics card. With one, it gets even faster.

Cloud AI? Only if you want it.

If you like, you can connect an online AI service with your own API key. It is off by default, and then everything stays on your PC.

Setup

Done before your coffee is.

Download, confirm once in Windows, and the setup assistant guides you through the rest. With short explainer videos from me (currently in German) and a test dictation at the end. You don’t need admin rights.

The video shows the German version. On an English Windows, FlowScribe starts in English.

Built in Wesel, Germany

“My voice was going to other people’s servers. So I built FlowScribe myself.”

I dictate every day: emails, quotes, messages. FlowScribe runs entirely on your PC, even without a graphics card, in any app. Now I’m looking for people who use it in their daily work and tell me honestly what’s missing.

Manuel LandeckDeveloper of FlowScribe · Firmenflow, Wesel

Manuel Landeck sits at his desk with a laptop, checking a quote
Manuel Landeck in a blazer in the park, recording a note by voice
Portrait of Manuel Landeck in a blazer, smiling while dictating an email
Manuel Landeck in a light linen shirt outside his front door, speaking a short message
Firmenflow logo

FlowScribe is software by Firmenflow.

Firmenflow is my business in Wesel, Germany. I build websites and look after Google Business Profiles for local businesses on the Lower Rhine and online. FlowScribe exists because I dictate every day myself.

Visit firmenflow.de (German) ›

FAQ

Quick answers.

What you should know before you install FlowScribe.

What does FlowScribe cost?

The beta is free, with no account and no sign-up. Costs only come up if you choose to use cloud AI with your own API key. The provider then bills you directly.

What kind of PC do I need?

Windows 10 or 11, 64-bit. You don’t need a graphics card: during setup, FlowScribe checks your processor, memory and graphics card and configures itself to match. With a dedicated graphics card with 4 GB of video memory or more, everything runs on it and is especially fast. On the processor alone, speech recognition is still quick, while AI editing takes a few seconds. Depending on your choice, the models need about 1 to 4 GB of free disk space. FlowScribe is not available for Mac or phones.

How do the models get onto my PC?

The setup assistant downloads them once: the speech recognition model (about 0.7 to 0.9 GB) and, if you want, an AI model for text editing (Qwen3.5, 1.3 or 2.7 GB). Both come from Hugging Face, a well-known platform for AI models. After that, FlowScribe also works offline.

Does anything really go online?

Your dictations stay on your PC. FlowScribe only stores recordings if you turn that on yourself. The app makes just three kinds of connections: model downloads when you load a model, an update check on GitHub at most once a day (you can turn it off), and cloud AI if you switch it on yourself. No usage data or crash reports are sent. You can check this in the app under Settings › Privacy › Network overview.

What happens if I turn on cloud AI?

It is off by default. You can turn it on under Settings › AI with your own API key for OpenAI, Anthropic (Claude), Mistral, OpenRouter, Groq or a service of your own. Then only the recognized text and your instruction go to that provider, never the audio.

Which languages does FlowScribe understand?

The recommended models understand German and English, even mixed in one sentence. For other languages, choose the “Many languages” model: it recognizes 25 European languages, such as French, Spanish or Polish. The app itself is available in English and German and follows your Windows language. You can change it in the settings.

Which programs does it work in?

Anywhere you can type: Outlook, Word, WhatsApp, your browser or ChatGPT. The exception is programs running as administrator. Windows doesn’t let other programs insert text there.

Why does Windows show a warning on first launch?

FlowScribe is new, so Microsoft doesn’t know the file yet. Click More info and then Run anyway. If you’d like to check first, checksums and VirusTotal reports for every version are on GitHub.

Do I need admin rights?

No, neither for installing nor for updates. For PCs where you’re not allowed to install anything, there is a portable version as a ZIP file.

How do I get updates?

Automatically. FlowScribe lets you know as soon as a new version is out. One click installs it, then the app restarts. Settings, history and models are kept. You update the portable version by hand.

How do I uninstall FlowScribe?

Like any program, under Windows Settings › Apps › Installed apps. If you tick “Also delete my settings and history”, your settings and history are removed. Without the tick, they stay for a later reinstall. Either way, the downloaded speech models remain in the shared Hugging Face folder (.cache\huggingface in your user folder), where you can delete them by hand.

What’s new

New in 0.9.6.

What changed with the latest updates. FlowScribe tells you when a new version is out, one click installs it.

NEWVersion 0.9.6November 2026

  • See your text as you speak

    While you dictate, FlowScribe shows what it has understood so far above the recording indicator, about once a second, with every recognition model. The finished text is inserted as before. Turn it on or off under Settings › Recording.

  • Inbox folder for voice messages

    Pick a folder under Import audio, for example in OneDrive. Every voice message that lands there turns into text on its own and then moves to “Done”.

  • Phone guide

    Step by step: how to get WhatsApp voice messages from an iPhone or Android phone onto your PC, including via WhatsApp Desktop and WhatsApp Web.

  • New paragraph by voice

    Say “new paragraph” and a new paragraph starts right there, in every mode. Paragraph numbers in a contract stay in the text as they are.

  • Voice messages with paragraphs

    On import, longer pauses and topic changes become paragraphs. The AI polishes each paragraph on its own instead of one big block.

Earlier versions

Version 0.9.4November 2026
  • Import audio as its own page: drop a file in or pick one. The import keeps running when you switch pages.
  • A short tour of every menu item, available again any time under Help.
  • New app icon, the same logo as on this website.

Version 0.9.3November 2026
  • The AI for polished text has its own step in setup.
  • The recognition model loads right after setup, even if its download was still running.

Version 0.9.2November 2026
  • Feedback box in the sidebar, one click opens a short form.
  • All texts rewritten in plain language, in German and English.

Version 0.9.1November 2026
  • “Optimized” no longer turns long dictations into bullet points.
  • Pasting without stalls, no doubled word fragments at the start of a dictation.
  • Short explainer videos in the setup assistant, even offline.

Version 0.9.0October 2026
  • The first beta: dictation in any app, modes like Email, Chat and Optimized.
  • Dictionary, snippets, history and stats.
  • Voice message import and automatic updates.

Try it in your daily work.

Try it here first. Then use it for a few days with emails, chats and reports, and tell me what’s missing.

NotepadExample

Hold, “speak”, let go. Just a simulation, nothing is recorded.

Try the free beta

Windows warns you on first launch because FlowScribe is new: More info › Run anyway.

FlowScribe runs on Windows 10 and 11. Open this page on your PC or send yourself the link.

Version 0.9.6 · Windows 10 and 11 · free

Portable version (ZIP, no installation) · Checksums, VirusTotal and all versions on GitHub

Like FlowScribe?

I build and maintain FlowScribe myself. If the app makes your day easier, you’re welcome to buy me a coffee.

Buy me a coffee (PayPal, opens in a new tab)