User guide
The same guide that ships with the app, for Mac and Windows. Inside VoxaFlow: Help › Guide …
VoxaFlow is a dictation app for Mac and Windows. Hold a key, speak and release. VoxaFlow types the cleaned-up text wherever your cursor is, in any app. Speech recognition and text cleanup run entirely on your computer.
The same content is available in the app under Help › Guide …, with practice fields and guided practices. What applies only to the Mac or only to Windows is marked as such.
Contents
- Getting started
- Dictation
- Hands-free mode
- Command mode
- Corrections and formatting
- Dictionary
- Snippets
- Style per app
- Settings
- Medical package
- Legal & Tax package
- Coding package
- Privacy
- Troubleshooting
Getting started
VoxaFlow turns speech into clean text and inserts it wherever your cursor is. It works in any app where you can type: Mail, Messages, Slack, Word, the browser or the terminal.
Requirements
Mac
- Mac with Apple silicon (M1 or later) and macOS 14 or later
- about 3 GB of free storage for the models
- 16 GB of memory recommended; on Macs with 8 GB, VoxaFlow picks the smaller language model by itself
VoxaFlow is available at voxaflow.ai/download. The app is signed with Developer ID and notarized by Apple, and updates arrive automatically.
Windows
- Windows 11, 64-bit (Windows 10 is not supported)
- about 3 GB of free storage for the models
- at least 8 GB of memory for the language model, 16 GB recommended; VoxaFlow picks the right model for your PC by itself
VoxaFlow is available at voxaflow.ai/download. The installer is signed (LOGIN SystemHaus GmbH), and updates arrive automatically.
Install and set up
Mac
- Open the DMG and in the installer window drag VoxaFlow onto the Applications folder, as the arrow shows.
- Start VoxaFlow from the Applications folder. The welcome window guides you through setup: allow access to the microphone and Accessibility, choose and test the dictation key, wait until the models are loaded.
- The last step has “Open VoxaFlow at login”, switched on by default. That way the dictation key is ready right after every restart. You can change this under Settings › General.
After that VoxaFlow lives as an icon in the menu bar. It has a Dock icon only while the VoxaFlow window is open.
Windows
- Run
VoxaFlow.App-win-Setup.exe. You don’t need administrator rights. If Windows shows “Windows protected your PC”, click “More info” and then “Run anyway”. - After the installation, setup opens: choose the microphone and check the level, choose and test the dictation key, wait until the models are loaded.
- In the last step you try dictating and switch on “Open VoxaFlow at sign-in”. That way the dictation key is ready right after every restart. You can change this under Settings › General.
After that VoxaFlow lives as an icon in the notification area of the taskbar, to the right next to the clock. If you don’t see it there, click the small arrow next to it. Clicking the icon opens the VoxaFlow window.
How it works
- Click into a text field.
- Hold
fn/Right Ctrl keyand speak. - Release
fn/Right Ctrl key. After about a second the text appears.
While you speak, a small pill with a level meter appears at the bottom of the screen. When you see it, VoxaFlow is listening.
Said: um so I just wanted to say that I’ll be a bit late today
Written: I just wanted to say that I’ll be a bit late today.
Mac
VoxaFlow does not just transcribe. A language model on your Mac (VoxaFlow Lektor, or VoxaFlow Lektor Kompakt on Macs with 8 GB of memory) removes filler words, adds punctuation and capitalization, and keeps only the corrected version when you change your mind mid-sentence.
Everything runs locally on this Mac. Neither audio nor text is sent to a server.
Windows
VoxaFlow does not just transcribe. A language model on your PC (VoxaFlow Lektor, or VoxaFlow Lektor Kompakt on PCs with less memory or an older processor) removes filler words, adds punctuation and capitalization, and keeps only the corrected version when you change your mind mid-sentence.
Everything runs locally on this PC. Neither audio nor text is sent to a server.
Dictation
Hold fn / Right Ctrl key while you speak. As soon as you release it, VoxaFlow processes the recording and inserts the text.
What VoxaFlow improves
- Filler words such as “um”, “uh” or “like” are removed.
- Punctuation and capitalization are added.
- When you correct yourself, only the corrected version remains.
- Numbers, times and amounts are written the way they are usually written.
- Questions are written as questions, not answered.
Said: hi sarah um thanks for your email I’ll get back to you by friday
Written: Hi Sarah, thanks for your email. I’ll get back to you by Friday.
Said: how much does the license for ten users cost
Written: How much does the license for ten users cost?
Which language model?
Mac
Under Settings › AI › Cleanup › Model you choose which model cleans up the text. VoxaFlow Lektor is recommended, and VoxaFlow Lektor Kompakt on Macs with 8 GB of memory. Both are built on Qwen3 and trained for dictation. The large model is more accurate but needs about 2.3 GB of memory. If memory runs short, macOS swaps to disk and every dictation gets noticeably slower. In tests Apple Intelligence was slower and less accurate than both Lektor models.
| Model | Short dictation | In tests |
|---|---|---|
| VoxaFlow Lektor, local | 0.2–0.6 s | most accurate |
| VoxaFlow Lektor Kompakt, local | about half as long as VoxaFlow Lektor | as accurate in everyday dictations, slightly weaker with medical and legal terms |
| Apple Intelligence | 0.6–2 s | considerably less accurate |
| Off (rules only) | instant | no corrections, lists or style |
Measured on a Mac with M4 Pro. Times may differ on other Macs. Long dictations take longer with every model because more text is generated. In Low Power Mode cleanup takes about twice as long.
Windows
Under Settings › AI › Cleanup › Model you choose which model cleans up the text. The default is Automatic: VoxaFlow uses VoxaFlow Lektor if your PC has enough memory and a fast processor, otherwise VoxaFlow Lektor Kompakt. Both are built on Qwen3 and trained for dictation. Which model is chosen for your PC and why is shown right below. The large model is more accurate, the small one faster. With less than 8 GB of memory VoxaFlow works with rules only.
| Model | In tests |
|---|---|
| VoxaFlow Lektor, local | most accurate |
| VoxaFlow Lektor Kompakt, local | as accurate in everyday dictations, slightly weaker with medical and legal terms |
| Off (rules only) | no corrections, lists or style |
Cleanup runs on the processor. How long a dictation takes therefore depends a lot on the PC. Long dictations take longer because more text is generated. In energy saver mode and on battery, cleanup can take considerably longer.
Long dictations
With longer dictations VoxaFlow works while you speak: from about 8 seconds on, finished sentences are recognized and cleaned up in the background. After you release the key only the last part is left.
Mac
So even a minute of dictation usually appears after about a second.
Windows
So even a minute of dictation usually appears after a few seconds.
You don’t need to do anything or pause for this.
If you correct yourself across a sentence end (“… on Monday. No, on Tuesday.”) or continue an enumeration (“… and secondly …”), VoxaFlow cleans up those sentences together.
Cancel
Press Esc during the recording and nothing is inserted. In hands-free mode Esc pauses first, see “Hands-free mode”.
The pill at the bottom of the screen
- Red with level meter: VoxaFlow is listening.
- “Writing …”: the text is being cleaned up.
Mac
- Orange notice: the fn key is also assigned to another function. See “Troubleshooting”.
Windows
- Orange notice: something did not work, for example because there was no text field or the program runs as administrator. The pill and the chapter “Troubleshooting” tell you what to do.
Very short utterances of a few words are processed without the language model. That is faster, and with so little text the result is practically the same.
Mac
Notes
Under Notes in the main window you write notes or dictate into them with Dictate or fn. With iCloud sync you have the same notes on all your devices with VoxaFlow.
With Pro, Summarize sums a note up in a few sentences and bullet points. Apple Intelligence does this on your Mac (macOS 26 or later); nothing leaves your device. Share sends the note, the summary or both. With the Medical package on, VoxaFlow does not summarize notes.
Windows
Notes
Under Notes in the main window you write notes or dictate into them with Dictate. Notes can be searched, copied and sent on with Share. They stay on this PC and are part of the backup.
Hands-free mode
For longer texts you don’t have to hold fn / Right Ctrl key the whole time.
- Tap
fn/Right Ctrl keytwice in quick succession. - Speak as long as you like. The pill shows “Hands-free · press key to stop”.
- Tap
fn/Right Ctrl keyonce to stop the recording.
Said: before the meeting I need to first review the offers and second confirm the appointment with the client
Written: Before the meeting I need to:
- review the offers
- confirm the appointment with the client
Pause: Esc pauses a hands-free dictation instead of cancelling it. The pill shows “Paused”. fn / Right Ctrl key continues, Esc again discards the dictation. What you say during the pause is not recorded, and the pause does not count towards the length of the dictation. After ten minutes the pill reminds you once; nothing is discarded on its own.
With “One tap starts hands-free” under Settings › Keyboard, a short tap on fn / Right Ctrl key is enough, without a second one. Holding still dictates.
Mac
If Apple’s own dictation is also set to “Press 🌐 twice”, both start at the same time. VoxaFlow warns you about this under Settings › Keyboard.
With Pro you can pick an own hands-free key under Settings › Keyboard › Keys per mode, for example the right ⌥ Option. Tap once to start, tap again to finish. Tapping fn twice keeps working.
Windows
Under Settings › Keyboard › Keys per mode you can pick an own hands-free key, for example the right Shift key. Tap once to start, tap again to finish. Tapping Right Ctrl key twice keeps working.
Command mode
Command mode lets you rewrite existing text with a spoken instruction.
- Select the text you want to change.
- Hold
⌃ Control/Shiftandfn/Right Ctrl key. The order doesn’t matter: if you pressfn/Right Ctrl keyfirst, you have one second to add⌃ Control/Shift. No other keys are needed. - Say what should happen to the text and release.
VoxaFlow replaces the selection with the rewritten version. The pill turns purple and shows “Command”.
Example instructions
- “Make this shorter.”
- “Make it friendlier.”
- “Translate this into German.”
- “Turn this into a bulleted list.”
- “Fix the spelling.”
Said: Selected: “I unfortunately can’t make it tomorrow because I have another appointment that was scheduled a while ago.” Instruction: make this shorter
Written: Sorry, I can’t make it tomorrow, I already have an appointment.
If the text stays the same, VoxaFlow replaces nothing and the pill shows “Text unchanged”. Then say more precisely what should change, for example “say this in other words”.
Mac
If VoxaFlow finds no selection, it inserts nothing and the pill shows “No selected text found”. This way your instruction never replaces the selected text by mistake. In browsers VoxaFlow reads the selection through the clipboard if needed and puts your previous clipboard content back afterwards.
Command mode is part of Pro; it is free until January 31, 2027 and during the trial. With Pro you can also pick an own command key under Settings › Keyboard › Keys per mode. Hold it while you speak the instruction.
Windows
If VoxaFlow finds no selection, it inserts nothing and the pill shows “No selection found”. This way your instruction never replaces the selected text by mistake. VoxaFlow briefly reads the selection through the clipboard (Ctrl+C) and puts your previous clipboard content back afterwards. Command mode does not work in Remote Desktop and Citrix.
Under Settings › Keyboard › Keys per mode you can also pick an own command key. Hold it while you speak the instruction.
Corrections and formatting
You can speak the way you think. VoxaFlow tidies up the text afterwards.
Self-corrections
When you correct yourself, VoxaFlow keeps only the final version. Signal words are, for example, “no”, “sorry”, “I mean” or “actually”.
Said: let’s meet on thursday at 3 pm no sorry at 4 pm
Written: Let’s meet on Thursday at 4 pm.
Lists
If you enumerate with “first, second, third”, a numbered list is created.
Said: for the project we need first a logo second a website and third a flyer
Written: For the project we need:
- a logo
- a website
- a flyer
E-mail addresses
Say the address the way you would speak it: “Müller at abc punkt com” becomes mueller@abc.com. Instead of “at” you can also say “ät” or “Klammeraffe”. VoxaFlow writes umlauts and ß as ae, oe, ue and ss, because addresses with umlauts are hardly ever delivered. Addresses you need often are most reliable as a snippet.
Quotation marks
Say “quote” before and “unquote” after the quotation, or “open quote” and “close quote”. “He said quote I’ll be there unquote” becomes: He said “I’ll be there”. In German say „Anführungszeichen … Anführungszeichen“, „Anführungszeichen auf … zu“ or „Zitat Anfang … Zitat Ende“ and get „…“. Remote Desktop and Citrix get straight quotation marks. A single „in Anführungszeichen“ without a second one stays a word.
Headings
With the Medical or Legal & Tax package you set headings by voice: at the start of a sentence say „Abschnitt Befund“, „Überschrift Diagnose“ or „Procedere Doppelpunkt“. Without one of these packages such words stay ordinary text. Which names count and how the headings arrive in Word, Mail or practice software is in the chapter “Style per app” under “Headings”.
Paragraphs and punctuation
- “New paragraph” inserts a blank line.
- “New line” starts a new line.
- “Period”, “comma” and “question mark” at the end of a phrase become the character.
Usually this isn’t necessary: VoxaFlow adds punctuation and paragraphs on its own.
Dictionary
Speech recognition sometimes mishears names, technical terms and abbreviations. The dictionary defines how a term is spelled.
Add a term
Mac
- Open Settings › Dictionary and click
+at the bottom left. - Enter the term with the correct spelling, for example “VoxaFlow”.
- Under “Misheard as”, enter typical mistakes separated by commas, for example “Voxa Flow, Box a Flow”.
Windows
- In the VoxaFlow window, open the Dictionary page and click “Add”.
- Enter the term with the correct spelling, for example “VoxaFlow”.
- Under “Misheard as”, enter typical mistakes separated by commas, for example “Voxa Flow, Box a Flow”.
VoxaFlow already knows its own name: if the recognizer hears something like “Vuxaflo”, the text still says “VoxaFlow”. You don’t need an entry for it, and it doesn’t count towards your terms.
Said: I’m testing box a flow on my laptop
Written: I’m testing VoxaFlow on my laptop.
The misheard variants are replaced directly. In addition, the language model receives the terms and also recognizes similar-sounding variants.
Useful for customer and product names, colleagues with unusual names, and abbreviations such as “GDPR” or “KPI”.
Edit, import, export
Double-click a term to edit it. If you add a term that already exists, VoxaFlow merges both. “Import and export” at the top right saves the dictionary as JSON or as a CSV file for Excel (columns “Term;Misheard as”). Before an import you see how many terms are new, changed or the same, and choose whether the file adds to or replaces the dictionary.
Suggestions from your corrections
When you correct a word after inserting, e.g. “MLS” to “MLX”, VoxaFlow remembers it. This works for up to five minutes after inserting and until the next dictation, even if you switch to another app in between. Suggestions then appear at the top of the dictionary; accept them with “Add” or ignore them. Only the text field VoxaFlow just wrote into is read, never a password field. You can turn this off under Settings › Inserting.
Faster with the shortcut “Add selection to dictionary”: select the wrong word, press the shortcut, type the correct spelling.
Free and Pro
Mac
Free applies the first 25 terms. You can add more; they stay saved, appear grey with a lock and work with Pro. Import and export are part of Pro; the backup under iCloud & Data always works.
Windows
Until January 31, 2027 all terms apply on Windows, import and export included; after that Free uses the first 25 terms, and import and export come with Pro.
Snippets
Snippets are text blocks. You say a short trigger and VoxaFlow inserts the stored text.
Add a snippet
Mac
- Open Settings › Snippets and click
+at the bottom left. - Enter a trigger, for example “my address”.
- Enter the text to insert.
Windows
- In the VoxaFlow window, open the Snippets page and click “Add”.
- Enter a trigger, for example “my address”.
- Enter the text to insert.
Said: my address
Written: 12 Example Street, Springfield 12345
Said: my calendar link
Written: https://cal.example.com/meeting
Snippets are inserted verbatim, without rewriting by the language model. Links, addresses and wording stay exactly as you stored them.
Pick triggers you would not say by accident in normal speech.
Edit, import, export
Double-click a snippet to edit it. “Import and export” saves your snippets as JSON or CSV (columns “Trigger;Text”) to load them on another computer.
Mac
With iCloud sync this happens automatically between your Macs.
Free and Pro
Mac
Snippets are part of Pro. Without Pro your snippets stay saved and work again as soon as Pro is active.
Windows
Until January 31, 2027 snippets are free on Windows. After that they are part of Pro; your snippets then stay saved and work as soon as Pro is active.
Style per app
VoxaFlow detects which app you are writing in and adapts the tone. Under Settings › AI › Style per app you choose how text should sound for each category.
Mac
| Category | Examples | Default |
|---|---|---|
| Mail, Outlook, Spark | Formal | |
| Chat | Slack, Messages, WhatsApp, Teams | Casual |
| Code & terminal | Xcode, VS Code, Terminal | Casual, technical terms stay unchanged |
| Documents & notes | Notes, Pages, Word, Notion | Casual |
| All other apps | Casual | |
| Practice software (Medical package) | Apps you assign under Packages; Remote Desktop, Citrix | Findings and notes |
| Law and tax office software (Legal & Tax package) | Apps you assign under Packages; Remote Desktop, Citrix | Brief and client letter |
Windows
| Category | Examples | Default |
|---|---|---|
| Outlook, Thunderbird | Formal | |
| Chat | Teams, Slack, WhatsApp | Casual |
| Code & terminal | VS Code, Visual Studio, Windows Terminal | Casual, technical terms stay unchanged |
| Documents & notes | Word, OneNote, Notion | Casual |
| All other apps | Casual | |
| Practice software (Medical package) | Apps you assign under Packages; Remote Desktop, Citrix | Findings and notes |
| Law and tax office software (Legal & Tax package) | Apps you assign under Packages; Remote Desktop, Citrix | Brief and client letter |
The same sentence in different apps
Said: yeah works for me I’ll be there at three
Written: Email (formal): Yes, that works for me. I’ll be there at three o’clock.
Said: yeah works for me I’ll be there at three
Written: Chat (very casual): yeah works for me, I’ll be there at three
Email format
In emails VoxaFlow puts your spoken greeting on its own line with a blank line after it, splits the text into paragraphs and puts your spoken closing with your name on their own lines at the end. Greeting and closing only appear if you said them; if you dictate a single paragraph into an existing email, it stays that paragraph. This works in email programs such as Mail, Outlook, Spark or Thunderbird and also in webmail in the browser, such as Gmail, Outlook on the web, GMX or WEB.DE. VoxaFlow recognizes webmail by the window title; the title is only compared, never stored.
Said: hi tom thanks for the update the numbers look good to me could you send the final version to the team by thursday best regards holger
Written: Hi Tom,Thanks for the update. The numbers look good to me. Could you send the final version to the team by Thursday?
Best regards,
Holger
Subject: If you start a German email with „Betreff“, the rest up to the first full stop becomes the first line, then a blank line, then the email as usual. „Betreff Angebot Mai Punkt Hallo Herr Wagner, …“ becomes “Angebot Mai” with “Hallo Herr Wagner,” below it. Copy the first line into the subject field of your mail program.
To lay out text you already wrote, use command mode: select it and say “format this as an email”. You can switch the email format off under Settings › AI › Style per app.
Style changes the tone, not the content. VoxaFlow does not add greetings or sentences you did not say.
Headings (Medical, Legal & Tax)
With Medical or Legal & Tax you say “Abschnitt Befund”, “Überschrift Procedere” or “Diagnose Doppelpunkt” at the start of a sentence, and VoxaFlow sets the heading. Only fixed names count, such as Anamnese, Befund, Diagnose, Beurteilung, Therapie, Procedere, Verlauf, or Sachverhalt and Begründung; “Der Befund ist unauffällig” stays a sentence. Each section is cleaned up on its own, nothing is moved between sections.
Mac

To dictate a letter with sections:
- Click into the document, for example in Word, and hold the dictation key or start hands-free mode.
- Speak the text in one go and start each section with “Abschnitt …”, “Überschrift …” or the name followed by “Doppelpunkt”. You do not need to say the period after it.
- Release the key. VoxaFlow sets the headings, writes units and dosages by the writing rules and inserts everything at once.
Said: abschnitt anamnese seit drei tagen husten und fieber keine thoraxschmerzen befund doppelpunkt blutdruck hundertvierzig zu neunzig abschnitt procedere ibuprofen vierhundert milligramm eins null eins
Written: Practice software:
ANAMNESE:
Seit drei Tagen Husten und Fieber, keine Thoraxschmerzen.BEFUND:
Blutdruck 140/90 mmHg.PROCEDERE:
Ibuprofen 400 mg 1-0-1.
Said: der befund ist unauffällig
Written: Der Befund ist unauffällig.
How headings arrive is set per category under “More …”:
Mac
- Formatted (Email, Documents & notes): bold, in the document’s font, for example in Word, Pages, Mail and Outlook
Windows
- Formatted (Email, Documents & notes): bold, in the document’s font, for example in Word and Outlook
- Plain text, headings in capitals (practice and office software): “BEFUND:” on a line of its own
- Plain text (all other apps): “Befund:” on a line of its own
Remote Desktop and Citrix always receive plain text.
Output language and own instruction
With Pro each category has “More …”:
- Output language: “As spoken” keeps your language. If you choose German or English, VoxaFlow translates into it, for example speak German and write English in chat. If the translation does not work out, VoxaFlow inserts the text as spoken and says so in the pill.
- Own instruction: up to 300 characters that add to the style. Instructions about form and closing work well, for example “End every email with: Best regards, Anna”, “no exclamation marks” or “all lower case”. Rewording the text, such as switching from formal to informal address, is beyond the local model.
Said: ich schicke dir die unterlagen äh morgen früh per mail
Written: Chat, output language English: I’ll send you the documents by email tomorrow morning.
Mac
Without Pro your settings stay saved but have no effect.
Windows
Until January 31, 2027 both are free on Windows.
Settings
Mac
Open the VoxaFlow window from the VoxaFlow icon in the menu bar › Open VoxaFlow … or with ⌘ ,. It starts with the overview. On the left is a sidebar like in System Settings. The search field at the top finds a setting even if you don’t know its section, for example with “clipboard”. While the window is open, VoxaFlow shows in the Dock and when switching with ⌘ ⇥.
| Section | What you find there |
|---|---|
| Overview | words, time saved, dictations, speaking speed and corrections |
| History | your dictations with search, copy and cleaning up again |
| General | app language, open at login |
| Keyboard | dictation key, key check and extra shortcuts |
| Inserting | clipboard and learning from your corrections |
| Audio | microphone, fallback microphones, sound output, muting and sounds |
| AI | dictation language, language model, style per app |
| Dictionary | terms, typical misrecognitions and suggestions |
| Snippets | text snippets |
| iCloud & Data | iCloud sync, retention, statistics and backup |
| Permissions | microphone and Accessibility |
| Packages | Medical, Legal & Tax and Coding: word lists, writing rules, practice and office software, professional secrecy, Remote Desktop |
| Coding | projects, terminal and English prompts (with the Coding package) |
| License | license key, packages, devices and buying |
| About | version, updates, guide and models used |
Windows
Open the VoxaFlow window by clicking the VoxaFlow icon in the notification area of the taskbar or from its menu › Open VoxaFlow …. It starts with the overview. The sections are on the left.
| Section | What you find there |
|---|---|
| Overview | words, time saved, dictations and corrections |
| History | your dictations with search, copy, cleaning up again, export and delete |
| Notes | write or dictate notes, search, copy and share them |
| General | app language, open at sign-in, short how-to |
| Keyboard | dictation key, keys per mode, shortcuts and key check |
| Insert | clipboard and learning from your corrections |
| Audio | microphone, fallback microphones, sound output, lowering other playback and sounds |
| AI | dictation language, language model, acceleration, style per app |
| Dictionary | terms, typical misrecognitions and suggestions |
| Snippets | text snippets |
| Packages | Medical and Legal & Tax: word lists, writing rules, practice and office software, professional secrecy, Remote Desktop and Citrix |
| Coding | projects, word lists, terminal, English prompts and spoken casing |
| Data | history, retention, statistics and backup |
| License | license key, packages, devices and purchase |
| About | version, updates, guide, feedback and models used |
Many settings have a small question mark next to them. Clicking it explains the setting with an example and leads to the matching chapter of this guide.
Overview
The overview counts how many words you dictate and how much time you save compared with typing. It assumes a typing speed of 40 words per minute.
Mac
Change it right in the “Time saved” tile.
Windows
You change it under Settings › Data.
The time you spend speaking and the processing are subtracted. “Corrections” shows separately what the AI cleanup, the dictionary and you yourself changed after inserting. Only numbers are counted, never the dictated text.
History
Mac
- Search and filter: the search field at the top searches the final text and the original. The filter limits the list to a period or an app.
- Copy: double-click an entry, press
⌘Cor use the copy icon that appears in the row on hover. “Copy” and “Copy original” are also on the right. - Clean up again: under “…” VoxaFlow sends the original through the AI once more, optionally in a different style.
- Several entries: select several with
⌘or⇧, copy them together or export them as text, CSV or JSON.
Windows
- Search and filter: the search field at the top searches the final text and the original. Next to it you limit the list to an app or a period.
- Copy: double-click an entry to copy it. “Copy” and “Copy original” are above the text.
- Export and delete: “Export …” saves the selected dictations as text, CSV or JSON. “Delete” or the
Delkey removes them from the history after asking. - Several entries: select several with
CtrlorShiftand copy, export or delete them together.
General
Mac
- App language: system language, Deutsch or English. The change takes effect after restarting VoxaFlow; a “Restart now” button appears for this.
- Open at login: starts VoxaFlow automatically. macOS then lists it under General › Login Items. During setup in the welcome window it is switched on by default.
Windows
- App language: Automatic (follows the Windows display language), Deutsch or English. The change takes effect immediately.
- Open VoxaFlow at sign-in: starts VoxaFlow automatically when you sign in to Windows. You switch it on in the last step of setup.
- How to dictate: the keys for dictating, hands-free and command mode at a glance.
Keyboard
Mac
- Dictation key:
fn, the right⌥ Optionor the right⌘ Command. - Key check: shows whether the key is also assigned to another function, such as Emoji & Symbols or Apple’s dictation. “Test key” checks whether the key reaches VoxaFlow at all.
- Shortcuts: set your own key combination for four actions. Click the field and press the combination;
⌫removes it.
| Action | What it does |
|---|---|
| Open quick history | window with search across recent dictations. ↩ inserts the selected text into the app you are writing in, ⌘ C copies it. |
| Paste last dictation again | inserts the newest dictation at the cursor once more |
| Process last recording again | recognizes the last recording again, e.g. after an error. The audio is only kept in memory for this. |
| Add selection to dictionary | select a misspelled word, press the shortcut and enter the correct spelling |
Windows
- Dictation key: the right Ctrl key (default), the right Shift key, Caps Lock, Scroll Lock or Pause. The right Alt key is missing because on German keyboards it is AltGr (@, €).
- Keys per mode: an own key for hands-free (one tap starts, another tap stops) and for command mode (hold it while you speak the instruction).
- Shortcuts: set your own key combination with Ctrl, Alt or Win, or an F key, for four actions. Click the field and press the combination; Backspace removes it,
Esccancels. - Key check: press a key and VoxaFlow shows whether it arrives.
| Action | What it does |
|---|---|
| Open quick history | window with search across recent dictations. Enter inserts the selected text into the app you are writing in, Ctrl Enter the original wording, Ctrl C copies it. |
| Paste last dictation again | inserts the newest dictation at the cursor once more |
| Process last recording again | recognizes the last recording again, e.g. after an error. The audio is only kept in memory for this. |
| Add selection to dictionary | select a misspelled word, press the shortcut and enter the correct spelling |
Inserting
- Restore clipboard after inserting: VoxaFlow inserts text via the clipboard and then puts your previous content back.
- Keep the text on the clipboard when there is no text field: if no text field is active, e.g. on the desktop, the text is not lost. The pill shows “No text field” and you paste with
⌘V/Ctrl+V. - Learn from my corrections: when you change a word right after inserting, VoxaFlow suggests it in the dictionary. More in the chapter “Dictionary”.
Audio
Mac
- Microphone and Sound output: “System default” follows the device currently set in macOS. A fixed device stays selected.
- Fallback microphones: if the chosen microphone is not connected, VoxaFlow tries the devices under “If it is not connected, try” in order, then the system default.
- Turn other audio down while recording: music, videos and notification sounds get quieter from the start sound until the recording ends. Afterwards the previous volume comes back. Devices without a volume control are muted instead. On by default.
- Ignore quiet voices in the background: speech that is much quieter than your own voice, such as someone talking a few metres away, is left out before recognition. Voices talking at the same time as you cannot be separated. On by default.
- Sounds: turns the start and end sounds on or off. Choose the start sound from the list and preview it with the speaker icon.
With AirPods it pays to choose the Mac’s microphone. Otherwise macOS switches the AirPods to headset mode and audio quality drops.
Windows
- Microphone and Sound output: “Windows default” follows the device currently set in Windows. A fixed device stays selected.
- Fallback microphones: if the chosen microphone is not connected, VoxaFlow tries the fallback microphones in order, then the Windows default. You change the order with the arrows.
- Lower other playback while recording: music and videos keep playing at 20 % during the recording. Afterwards the previous volume comes back. On by default.
- Ignore quiet voices in the background: speech that is much quieter than your own voice, such as someone talking a few metres away, is left out before recognition. Voices talking at the same time as you cannot be separated. On by default.
- Sounds: turns the start and end sounds on or off. Choose the start sound from the list and preview it with ▶.
With Bluetooth headphones it pays to choose the PC’s microphone. Otherwise Windows switches the headphones to headset mode and audio quality drops.
AI
- Dictation language: “Automatic” detects the language itself. A fixed language helps when short sentences are recognized in the wrong language.
Mac
- Model: VoxaFlow Lektor (most accurate), VoxaFlow Lektor Kompakt (faster, recommended on Macs with 8 GB of memory), Apple Intelligence, or “Off” with simple rules. The comparison is in the chapter “Dictation”.
- Style per app: Formal, Casual or Very casual, separately for email, chat, code & terminal, documents & notes and all other apps.
- Local models: shows whether speech recognition and the language model are loaded.
Windows
- Model: Automatic (recommended, matched to your PC), VoxaFlow Lektor (most accurate), VoxaFlow Lektor Kompakt (faster), or “Off” with simple rules. The comparison is in the chapter “Dictation”.
- Full performance on battery while dictating: sets the Windows power mode to “Best performance” during a dictation and back to your mode afterwards. On battery the AI then works about twice as fast. It stays off in energy saver. On by default.
- Acceleration: Automatic or processor only; with a graphics card that has enough memory of its own, also the graphics card.
- Style per app: Formal, Casual or Very casual, separately for email, chat, code & terminal, documents & notes and all other apps. “More …” holds the output language, your own instruction and, with Medical or Legal & Tax, how dictated headings appear.
- Local models: shows whether speech recognition and the language model are loaded.
Mac
iCloud & Data
- Sync with iCloud: dictionary, snippets, settings and statistics on all your Macs, and the history if you want. The data lives in the “VoxaFlow” folder in your iCloud Drive and is end-to-end encrypted. Only your devices have the key.
- Recovery code: the first Mac creates the key. On every further Mac you enter the recovery code you find under “Recovery code …”. Keep it safe, for example in your password manager.
- History: whether to save it, and for how long: 7 days, 30 days, 1 year or forever.
- Statistics: turns the numbers for the overview on or off.
- Backup: puts settings, styles, dictionary and snippets into one file, optionally with the history. When restoring you choose which parts come back.
Windows
Data
- Save history: yes or no; turning it off deletes the history.
- Keep dictations for: how long the history keeps dictations.
- Collect statistics for the overview: numbers only, no text.
- Your typing speed: for “Time saved” in the overview.
- Folder: where data, models and the log are stored.
- Backup: puts settings, styles, dictionary, snippets and notes into one file, optionally with the history, for example for a new PC. When restoring you choose which parts come back; dictionary, snippets, notes and history are merged. A backup from the Mac brings the dictionary, snippets, notes and history along.
Packages
Mac
The Packages page appears once your license includes a package; until January 31, 2027 that is all three. You switch each package on as a whole with “Use the … package”, all are off at first.
Windows
Until January 31, 2027 all packages are available. You switch on Medical and Legal & Tax on the Packages page with “Use the … package”; they are off at first. Coding has a page of its own and is on by default.
Below it you choose the word lists. VoxaFlow downloads a list the first time you switch it on and checks its signature. While you dictate, VoxaFlow looks in the active lists for terms you may have said, also when they sound alike. Your own dictionary comes first, and a word is never replaced just because it resembles a term.
- Medical and Legal & Tax add writing rules, headings by voice, a category of their own for practice or office software, professional secrecy mode and settings for Remote Desktop and Citrix. Explained step by step in the chapters “Medical package” and “Legal & Tax package”, each with its own practice under Help.
- Coding adds terminal commands, English prompts and spoken casing, described in the chapter “Coding package”.
The first time you switch on Medical or Legal & Tax, a short window shows the first steps. Later you find the same buttons at the top of the package under “Getting started”.
Mac
Remote Desktop and Citrix. VoxaFlow runs on your Mac, recognizes and cleans up there and inserts the text into the window of the Windows App (Remote Desktop) or Citrix Workspace; the application in the session receives it as if typed. For each client you choose:
- Typing (default, also without the packages): VoxaFlow types the text character by character. This also works when the session blocks the clipboard, and your clipboard stays untouched. Long texts take a few seconds.
- Clipboard: faster for long texts. VoxaFlow pastes with ⌘V as soon as the client has passed the clipboard on to the session. This only works if the session allows the clipboard and the Windows App is set to the Scancode keyboard mode (menu Connections › Keyboard Mode); in Unicode mode only a “v” arrives.
Windows
Remote Desktop and Citrix. VoxaFlow runs on your PC, recognizes and cleans up there and inserts the text into the window of Remote Desktop (including the Windows App) or Citrix Workspace; the application in the session receives it as if typed. For each client you choose:
- Type (default, also without the packages): VoxaFlow types the text character by character. This also works when the session blocks the clipboard, and your clipboard stays untouched. Long texts take a few seconds.
- Clipboard: faster for long texts. VoxaFlow pastes with
Ctrl+Vas soon as the client has passed the clipboard on to the session. This only works if the session allows the clipboard.
VoxaFlow cannot see which program is in front inside the session. So there is no “No text field” notice, no style per program of the session, no learning from corrections and no command mode for selected text there.
License
Mac
Until January 31, 2027 all features and all packages (Coding, Medical, Legal & Tax) are included, without a license key; the License page shows until when.
Launch price for early users. Everyone who uses VoxaFlow before February 1, 2027 gets the launch price on every purchase until February 28, 2027. Buying under License › Buy on this Mac recognizes the device by itself. For buying on another computer, the License page shows your early-user code (FRUEH-…); enter it there when you buy. We don’t need an email address for this. If you bought VoxaFlow, enter the license key from the purchase email here (VOXA-PRO-… or VOXA-TEAM-…; case and spaces do not matter) and click Activate. You then see your packages and how many devices use the license. It is a one-time purchase without a subscription; all updates of VoxaFlow 1 are included. A single license runs on 2 Macs, a practice and office license on 5. “Deactivate this Mac …” frees a seat for another Mac; you can also do this in your customer account on voxaflow.ai.
VoxaFlow checks the license every 30 days with license.voxaflow.ai, every 7 days until January 31, 2027. It only sends a checksum calculated from the hardware ID, the app version and the key or license token, never text or audio. Without internet everything keeps working; the license stays valid for 14 days beyond the check date. You can buy packages or a new license right on the page; the checkout opens in your browser.
Free and trial. From the start of sales on February 1, 2027, VoxaFlow starts on a new Mac with 14 days of Pro and all packages. After that, without a license, Free applies: 2,500 words a week (Monday to Sunday), 2 minutes per dictation, 25 dictionary terms and 7 days of history. The dictation that crosses the weekly limit is still inserted in full; after that the dictation key starts no recording until Monday. From 1:45 the pill counts down the seconds, at 2:00 the recording ends and the text is inserted. With Pro it is 15 minutes. The License page and the menu show how many words you dictated this week. The counter stays on your Mac and is never sent. These limits do not apply until January 31, 2027.
Windows
Until January 31, 2027 all features and all packages (Coding, Medical, Legal & Tax) are included on Windows, without a license key; the License page shows until when. If you have a license key, enter it there (VOXA-PRO-… or VOXA-TEAM-…) and click Activate. “Sign out on this PC” frees a seat for another device; you can also do this in your customer account on voxaflow.ai.
Launch price for early users. Everyone who uses VoxaFlow before February 1, 2027 gets the launch price on every purchase until February 28, 2027. Buying under License › Buy VoxaFlow on this PC, VoxaFlow recognizes the device by itself. For buying on another computer, the License page shows your early-user code (FRUEH-…) with a button to copy it; you enter it there when buying, on Windows in the field “Early-user code (optional)”. We don’t need an email address for that.
VoxaFlow regularly checks the license with license.voxaflow.ai. It only sends a checksum calculated from the hardware ID, the app version and the key or license token, never text or audio. Without internet everything keeps working.
Mac
Menu bar
The first line of the VoxaFlow menu bar menu shows whether VoxaFlow is ready or the models are still loading. Below are “Paste last dictation again” and “Copy last dictation”. The History submenu holds the last eight dictations (a click copies), the quick history, “Process last recording again”, “Add selection to dictionary …” and all dictations. Shortcuts you set under Settings › Keyboard appear next to the items. “Open VoxaFlow …” opens the VoxaFlow window at the top of the overview; the history, dictionary and all settings are in its sidebar. “Send Feedback …” is right in the menu and at the bottom of the sidebar of the VoxaFlow window. Help holds this guide, the guided practice, the practice of the Medical or Legal & Tax package while it is on, and “Check for Updates …”. If the dictation key is assigned twice, a permission is missing, an update is available or there are dictionary suggestions, an item appears that leads straight there. While an update is waiting, the menu bar icon shows a dot, and the update window comes up at most once a day, never during a dictation.
Windows
Notification area
When you hover over the VoxaFlow icon in the notification area of the taskbar, it shows whether VoxaFlow is ready or the models are still loading. A right-click opens the menu: the status at the top, below it “Paste last dictation again” and “Copy last dictation”. The History submenu holds the last eight dictations (a click copies), “Process last recording again”, “Add selection to dictionary …” and all dictations. “Open VoxaFlow …” opens the VoxaFlow window on the overview. Help holds “How to dictate …”, this guide, the guided practice, the practice of the Medical or Legal & Tax package while it is on, the guide on voxaflow.ai, “What’s new …”, the setup and “Send Feedback …”. While an update is waiting, the top of the menu shows “Install update to … (restart)” and below it “What’s new in …”; otherwise it is installed at the next start. A window shows what the update brings, at most once a day and never during a dictation. If you did not see it, VoxaFlow shows what’s new at the first start of the new version.
Medical package
With the Medical package VoxaFlow writes medical terms, drug names, values and dosages the way they belong in a doctor’s letter or a patient record. You dictate as usual into your practice software, into Word or into an email. Speech and text are processed only on your computer. It works in English (US and UK) and in German: VoxaFlow uses the word lists and writing rules of the language you dictate in.
Said: metoprolol fifty milligrams twice daily blood pressure one forty over ninety
Written: Metoprolol 50 mg twice daily, blood pressure 140/90 mmHg.
Set up in five steps
- Open VoxaFlow (icon in the menu bar / icon in the notification area › Open VoxaFlow …) and choose Packages on the left.
- Switch on Use the Medical package. The word lists appear below, for example general medicine, cardiology, drugs or nursing. Switch on the areas you work in. More does no harm; the first download just takes a moment longer.
- Under Practice software, click Add app … and choose your practice system, for example tomedo or Medistar. VoxaFlow then writes there in the style “Findings and notes”.
- Under English, choose US or British English. Automatic follows the region of your Mac. It decides spelling, units and drug names, for example acetaminophen or paracetamol.
- Do you work through Remote Desktop or Citrix? Then there is nothing more to do: VoxaFlow types the text into the session window. More below under “Remote Desktop and Citrix”.
Mac

Under Help › Medical practice … you try everything in three minutes with example sentences.
How to dictate
Click where the text should go. Hold fn / Right Ctrl key, speak and release it. The text appears after a moment.
- Speak naturally and in full sentences. You do not need to say punctuation.
- Simply say numbers, values and doses: “fifty milligrams”, “one forty over ninety”.
- Feel free to correct yourself mid-sentence: “five, sorry, ten milligrams” becomes “10 mg”.
- For longer dictations, tap
fn/Right Ctrl keytwice. Then you do not need to hold the key. Pressfn/Right Ctrl keyonce to end.
Values, doses and abbreviations
| You say | US English | British English |
|---|---|---|
| fifty milligrams | 50 mg | 50 mg |
| two point five milligrams | 2.5 mg | 2.5 mg |
| two hundred micrograms | 200 mcg | 200 micrograms |
| ten milliliters | 10 mL | 10 mL |
| BP one forty over ninety | BP 140/90 mmHg | BP 140/90 mmHg |
| ASA eighty one | ASA 81 | ASA 81 |
| one tablet b i d | one tablet BID | — |
| one tablet t d s | — | one tablet TDS |
| p r n | PRN | PRN |
| on the thirteenth of October | on October 13 | on 13 October |
Micrograms stay written out in British English, as the NHS recommends, so they are never mistaken for milligrams. Frequencies become abbreviations only when you spell them letter by letter; “twice daily” stays as you say it. A dash means: stays as you say it.
When you dictate in German, the German rules apply: “fünf Milligramm eins null eins” becomes “5 mg 1-0-1”, “Zustand nach” becomes “Z. n.”, “Verdacht auf” becomes “V. a.”.
VoxaFlow adds no numbers or dates you did not say and leaves none out. If the cleanup adds something anyway, VoxaFlow uses the text with the fixed rules only. You can switch these rules off under Packages › Writing rules.
Letters with sections
Start a section with “section” and its name, for example “section assessment”, or say the name followed by “colon”: “plan colon”. VoxaFlow sets the heading. Recognized names include History, Presenting complaint, Examination, Findings, Investigations, Assessment, Impression, Diagnosis, Plan, Medications, Allergies, Follow-up, Subjective and Objective. In German you say “Abschnitt Befund” or “Procedere Doppelpunkt”.
Mac

Said: section findings chest clear section plan review in one week
Written: Practice software:
FINDINGS:
Chest clear.PLAN:
Review in one week.
In Word and in emails the headings are bold, in practice software they appear in capitals. A sentence such as “The plan is to discharge tomorrow” stays an ordinary sentence.
When a word is spelled wrong
If VoxaFlow spells a name or a rare term wrong, add it to the dictionary: select the word and choose History › Add selection to dictionary … in the menu of the VoxaFlow icon. From then on VoxaFlow always writes it that way. To rework a finished text, use command mode: select the text, hold fn / Right Ctrl key together with ⌃ Control / Shift and say what should change, for example “make this more formal”.
Confidentiality and privacy
- Your speech and your text are recognized and cleaned up only on this computer. VoxaFlow sends neither audio nor text to the internet.
- As soon as the package is on, professional secrecy mode applies: new dictations are not kept in the history. If you want to keep them for a while, switch on Keep dictations in the history under Packages and choose 24 hours, 7 or 30 days. The history is never synced with other devices.
- Under Packages you save “What VoxaFlow sends” as PDF … for your record of processing activities. It lists which servers VoxaFlow contacts and why, for example for updates and the license.
Mac
If you are not bound by professional secrecy and only need the word lists, turn off Professional secrecy under Packages or under iCloud & Data. History and notes are then kept and synced as without the package. The setting applies to this Mac only.
Remote Desktop and Citrix
VoxaFlow runs on the computer you sit at and writes the finished text into the Remote Desktop or Citrix window. Your practice software in the session receives it as if you had typed it. This also works when the session blocks the clipboard. Long texts take a few seconds. Under Packages › Remote Desktop and Citrix you choose Typing or Clipboard for each client; clipboard is faster but needs a shared clipboard.
What VoxaFlow does not do
VoxaFlow writes what you dictate. It summarizes nothing, does not evaluate findings and suggests no diagnoses or codes. Check the text as you would a document you typed yourself.
Frequently asked questions
Do I need internet? Only once to download the models and word lists. After that you can dictate offline.
Does it work in my practice software? VoxaFlow writes into any program you can type into. If a program does not accept the text, it is on the clipboard and you paste it with ⌘V / Ctrl+V.
Do I have to speak very clearly? No. Speak normally, as with a dictation device. A good headset helps in noisy rooms.
What does the package cost? Until January 31, 2027 everything is free. From February 1, 2027, Pro costs €49 and the Medical package €40, a one-time purchase without a subscription. Practices get 5 seats for the price of 4.
Who helps me set it up? Write to us via “Send Feedback …” or at hello@voxaflow.ai. A guide with pictures is also at voxaflow.ai/docs/medical.
Legal & Tax package
With the Legal & Tax package VoxaFlow writes sections, case numbers, courts, legal terms and amounts the way they belong in a brief, a file note or a client letter. You dictate as usual into your office software, into Word or into an email. Speech and text are processed only on your computer. It works in English (US and UK) and in German: VoxaFlow uses the word lists and writing rules of the language you dictate in.
Said: the claim arises under forty two u s c section nineteen eighty three
Written: The claim arises under 42 U.S.C. § 1983.
Set up in five steps
- Open VoxaFlow (icon in the menu bar / icon in the notification area › Open VoxaFlow …) and choose Packages on the left.
- Switch on Use the Legal & Tax package. The word lists appear below, for example civil law, labor law, tax law, tax forms or Latin terms. Switch on the areas you work in. Each area has lists for US, UK and German law.
- Under Law and tax office software, click Add app … and choose your office program. VoxaFlow then writes there in the style “Brief and client letter”.
- Under English, choose US or British English. Automatic follows the region of your Mac. It decides which law the word lists and citations follow.
- Do you work through Remote Desktop or Citrix? Then there is nothing more to do: VoxaFlow types the text into the session window. More below under “Remote Desktop and Citrix”.
Mac

Under Help › Legal & Tax practice … you try everything in three minutes with example sentences.
How to dictate
Click where the text should go. Hold fn / Right Ctrl key, speak and release it. The text appears after a moment.
- Speak naturally and in full sentences. You do not need to say punctuation.
- Simply say citations and amounts: “forty two U S C section nineteen eighty three”, “twelve thousand five hundred dollars”.
- Feel free to correct yourself mid-sentence: “the deadline is the tenth, sorry, the twelfth of November”.
- For longer dictations, tap
fn/Right Ctrl keytwice. Then you do not need to hold the key. Pressfn/Right Ctrl keyonce to end.
Citations, amounts and dates
| You say | US English | British English |
|---|---|---|
| forty two U S C section nineteen eighty three | 42 U.S.C. § 1983 | — |
| fifteen U S C sections seventy eight and following | 15 U.S.C. §§ 78 et seq. | — |
| Rule twelve b six | Rule 12(b)(6) | — |
| twenty twenty four U K S C twelve | — | [2024] UKSC 12 |
| twenty twenty three E W C A civ forty five | — | [2023] EWCA Civ 45 |
| twelve thousand five hundred dollars | $12,500 | — |
| twelve thousand five hundred pounds | — | £12,500 |
| nineteen percent | 19% | 19% |
| on the thirteenth of October | on October 13 | on 13 October |
A dash means: stays as you say it. Acts stay as you say them (“section three of the Theft Act 1968”). When you dictate in German, the German rules apply: “Paragraf achthundertdreiundzwanzig Absatz eins BGB” becomes “§ 823 Abs. 1 BGB”, “Aktenzeichen eins B v R zwölf vierundzwanzig” becomes “Az. 1 BvR 12/24”.
VoxaFlow adds no numbers, dates or deadlines you did not say and leaves none out. “The deadline is the twenty-second” never becomes “October 22”. If the cleanup adds something anyway, VoxaFlow uses the text with the fixed rules only. You can switch these rules off under Packages › Writing rules.
Briefs and notes with sections
Start a section with “section” and its name, for example “section facts”, or say the name followed by “colon”: “conclusion colon”. VoxaFlow sets the heading. Recognized names include Facts, Background, Issues, Analysis, Discussion, Conclusion, Recommendation and Deadlines, in US English also Questions presented, Argument and Relief requested, in British English Submissions, Advice and Relief sought. In German you say “Abschnitt Sachverhalt” or “Ergebnis Doppelpunkt”.
Said: section facts the client resigned on the third of march section conclusion the dismissal is invalid
Written: Office software:
FACTS:
The client resigned on March 3.CONCLUSION:
The dismissal is invalid.
In Word and in emails the headings are bold, in office software they appear in capitals. A sentence such as “The conclusion is clear” stays an ordinary sentence.
When a word is spelled wrong
If VoxaFlow spells a name, a company or a rare term wrong, add it to the dictionary: select the word and choose History › Add selection to dictionary … in the menu of the VoxaFlow icon. From then on VoxaFlow always writes it that way. To rework a finished text, use command mode: select the text, hold fn / Right Ctrl key together with ⌃ Control / Shift and say what should change, for example “make this more formal”.
Confidentiality and privacy
- Your speech and your text are recognized and cleaned up only on this computer. VoxaFlow sends neither audio nor text to the internet.
- As soon as the package is on, professional secrecy mode applies: new dictations are not kept in the history. If you want to keep them for a while, switch on Keep dictations in the history under Packages and choose 24 hours, 7 or 30 days. The history is never synced with other devices.
- Under Packages you save “What VoxaFlow sends” as PDF … for your record of processing activities. It lists which servers VoxaFlow contacts and why, for example for updates and the license.
Mac
If you are not bound by professional secrecy and only need the word lists, turn off Professional secrecy under Packages or under iCloud & Data. History and notes are then kept and synced as without the package. The setting applies to this Mac only.
Remote Desktop and Citrix
VoxaFlow runs on the computer you sit at and writes the finished text into the Remote Desktop or Citrix window. Your office software in the session receives it as if you had typed it. This also works when the session blocks the clipboard. Long texts take a few seconds. Under Packages › Remote Desktop and Citrix you choose Typing or Clipboard for each client; clipboard is faster but needs a shared clipboard.
What VoxaFlow does not do
VoxaFlow writes what you dictate. It summarizes nothing, does not assess the legal situation and gives no legal or tax advice. Check the text as you would a document you typed yourself, especially deadlines and amounts.
Frequently asked questions
Do I need internet? Only once to download the models and word lists. After that you can dictate offline.
Does it work in my office software? VoxaFlow writes into any program you can type into. If a program does not accept the text, it is on the clipboard and you paste it with ⌘V / Ctrl+V.
Do I have to speak very clearly? No. Speak normally, as with a dictation device. A good headset helps in noisy rooms.
What does the package cost? Until January 31, 2027 everything is free. From February 1, 2027, Pro costs €49 and the Legal & Tax package €40, a one-time purchase without a subscription. Offices get 5 seats for the price of 4.
Who helps me set it up? Write to us via “Send Feedback …” or at hello@voxaflow.ai. A guide with pictures is also at voxaflow.ai/docs/legal.
Coding package
With the Coding package VoxaFlow knows the names in your project, writes terminal commands correctly and, if you like, English prompts. You dictate as usual into Cursor, VS Code, a terminal or a chat with an AI agent.
Said: refactor the fetch user profile hook so it uses the auth service
Written: Refactor the fetchUserProfile hook so it uses the AuthService.
Set up
Mac
- In the VoxaFlow window, choose Packages on the left and switch on Use the Coding package.
- Switch on the word lists of your tools, for example React, Swift, Python or Kubernetes.
- The Coding page appears on the left. It shows the detected projects and the settings for the terminal and English prompts.
Windows
- In the VoxaFlow window, choose Coding on the left. Use the Coding package is on by default.
- Switch on the word lists of your tools, for example React, C#, Python or Kubernetes.
- Below them you set up the terminal and English prompts.
Projects
Mac
When you dictate inside a git repository, VoxaFlow reads its file and folder names and its declared names, such as types and functions, once. “auth service” then becomes AuthService, “fetch user profile” becomes fetchUserProfile. Files ignored by git are skipped, and nothing of it leaves your Mac. Each project can have its own dictionary entries and snippets, which come before the global ones there.
Windows
When you dictate in a git repository, VoxaFlow reads its file and folder names and declared names such as types and functions once. “auth service” then becomes AuthService, “fetch user profile” becomes fetchUserProfile. Files ignored by git are skipped, and nothing of it leaves your computer. VoxaFlow finds the project in the window title: by the folder name in VS Code, Cursor and Windsurf, by the path in Command Prompt and Git Bash. For other programs, add the folder under Coding › Projects. Each project can have its own dictionary entries and snippets that come before the global ones there.
Terminal
When a dictation in a terminal starts with a command such as git, npm or docker, VoxaFlow writes it without AI.
Said: git commit minus m fix login bug
Written: git commit -m “fix login bug”
If an agent such as Claude Code runs in the terminal, or you say a normal sentence, you get a normal sentence.
Prompts in English
Can be switched on for editors and terminals separately. You speak German, VoxaFlow writes English and keeps names from the project, file names, paths and commands unchanged. If the translation changes one of them, the text comes as spoken.
Spoken casing
| You say | VoxaFlow writes |
|---|---|
| camel case user id | userId |
| pascal case user id | UserId |
| snake case user id | user_id |
| kebab case user id | user-id |
| constant max retries | MAX_RETRIES |
This works in every app.
Privacy
Mac
VoxaFlow works entirely on your Mac.
- Speech recognition (Parakeet) runs on your Mac’s Neural Engine.
- Text cleanup (VoxaFlow Lektor) runs on your Mac’s GPU.
- Audio and text are not sent to any server. After the one-time model download, VoxaFlow also works offline.
Windows
VoxaFlow works entirely on your PC.
- Speech recognition (Parakeet) runs on your PC’s processor.
- Text cleanup (VoxaFlow Lektor) runs on the processor or, if you have a graphics card with enough memory of its own, on the graphics card.
- Audio and text are not sent to any server. After the one-time model download, VoxaFlow also works offline.
What VoxaFlow sends to the internet
Never text or audio. All requests go to VoxaFlow’s own servers in Germany:
- Models: on first launch VoxaFlow downloads the models from
models.voxaflow.ai. Only the file itself is requested.
Mac
- Updates: once a day VoxaFlow asks
updates.voxaflow.aiwhether there is a new version. Only the app name and version are sent, no system profile. You can turn this off under Settings › About.
Windows
- Updates: once a day VoxaFlow asks
updates.voxaflow.aiwhether there is a new version. Only the app name and version are sent, not the Windows version. An update is downloaded in the background and installed at the next start; what it brings is part of the update feed and comes along.
- License:
license.voxaflow.aichecks the license (see Settings › License). - Word lists: if you switched on word lists under Packages, VoxaFlow downloads them from
packs.voxaflow.aiand looks for new versions at most once a day. Only the file is requested, without license data. - Feedback: only when you click Send under “Send Feedback …”.
The servers store IP addresses only shortened and delete their logs after 7 days.
What is stored
Mac
- History: your dictations are stored locally so you can copy them later. Under Settings › iCloud & Data you choose whether and how long it is kept, or delete it.
- Dictionary, snippets and settings: locally in your user folder.
- Statistics: numbers only (words, durations, apps), never the dictated text. You can turn them off under Settings › iCloud & Data.
- iCloud: only when you turn on sync, VoxaFlow stores the data in the “VoxaFlow” folder in your iCloud Drive. It is encrypted with AES-256 on your Mac first. Only your devices have the key (keychain and recovery code); neither Apple nor VoxaFlow can read the data. macOS does the transfer, VoxaFlow itself opens no connection for it.
- Professional secrecy: With the Medical or Legal & Tax package, VoxaFlow keeps dictations only if you allow it under Packages, for 30 days at most and on this Mac only. History and notes are then never synced via iCloud. If you are not bound by professional secrecy, you can turn it off under iCloud & Data.
- Clipboard: VoxaFlow inserts text via the clipboard and then restores its previous content. The inserted text is marked so clipboard managers ignore it.
Windows
- History: your dictations are stored locally so you can copy them later. Under Settings › Data you choose whether and how long it is kept, or delete it.
- Dictionary, snippets and settings: locally in your user folder under
%LOCALAPPDATA%\VoxaFlow. - Statistics: numbers only (words, durations, apps), never the dictated text. You can turn them off under Settings › Data.
- Professional secrecy: With the Medical or Legal & Tax package, VoxaFlow keeps dictations only if you allow it under Packages, for 30 days at most and on this PC only.
- Clipboard: VoxaFlow inserts text via the clipboard and then restores its previous content. The inserted text does not end up in the Windows clipboard history (
Win+V) and is not synced to your other devices.
Permissions
Mac
| Permission | Purpose |
|---|---|
| Microphone | to record your dictation |
| Accessibility | to detect the dictation key, insert text at the cursor and read selected text |
Windows
VoxaFlow does not need administrator rights on Windows. If microphone access is switched off for desktop apps, allow it under Windows Settings › Privacy & security › Microphone. VoxaFlow detects the dictation key through a Windows keyboard hook. It only evaluates the chosen keys and shortcuts and stores no input.
Troubleshooting
Nothing happens when I press the dictation key
Mac
- Open Settings › Permissions and check that both checkmarks are green (microphone and Accessibility).
- Open Settings › Keyboard, click “Test key” and press
fn. If “fn detected” appears, the key works. - If nothing arrives, your keyboard probably does not send fn to the Mac. Many external keyboards behave this way. Under Settings › Keyboard, set the dictation key to the right
⌥ Optioninstead.
Windows
- Check that VoxaFlow is running: its icon is in the notification area of the taskbar, possibly behind the small arrow.
- Open Settings › Keyboard and press
Right Ctrl keyunder “Key check”. If the key shows up there, it reaches VoxaFlow. - If nothing arrives, your keyboard or another program is using the key. Under Settings › Keyboard, choose a different dictation key, for example the right Shift key.
- If VoxaFlow listens but writes nothing, check under Windows Settings › Privacy & security › Microphone that desktop apps are allowed to access the microphone.
Microphone switched or headset disconnected
Mac
Under Settings › Audio you choose the microphone and the device for the sounds. “System default” follows the device currently set in macOS. If you choose a fixed device and it is not connected, VoxaFlow automatically uses the system default. When the device comes back, it is used again.
Windows
Under Settings › Audio you choose the microphone and the device for the sounds. “Windows default” follows the device currently set in Windows. If you choose a fixed device and it is not connected, VoxaFlow automatically uses the Windows default. When the device comes back, it is used again.
If you switch devices during a recording, VoxaFlow processes what was said so far. If a microphone delivers no audio after pressing the key, VoxaFlow restarts the recording once and otherwise shows “Microphone not responding”.
Mac
With AirPods it is worth choosing the Mac’s microphone. Otherwise macOS switches the AirPods to headset mode and audio quality drops.
fn opens emoji or Apple’s dictation
The 🌐 key is assigned twice. Open System Settings › Keyboard and set “Press 🌐 key to” to “Do Nothing”. If Keyboard › Dictation has the shortcut “Press 🌐 twice”, change it or turn Apple’s dictation off. VoxaFlow shows such conflicts in the settings and in the menu.
Permission granted, but VoxaFlow says it is missing
This happens when VoxaFlow was installed with a different signature, for example a development build instead of an official release. macOS ties the entries to the signature, not to the name. Under Settings › Permissions, click “Repair permissions”. VoxaFlow then removes the old entries and requests the permissions again. Switch VoxaFlow on again in System Settings under Accessibility.
Official releases and updates always have the same signature, so permissions are kept.
Windows
With Bluetooth headphones it is worth choosing the PC’s microphone. Otherwise Windows switches the headphones to headset mode and audio quality drops.
Windows shows a warning during installation
If Windows shows “Windows protected your PC”, click “More info” and then “Run anyway”. The installer is signed; the warning appears as long as a new version has rarely been downloaded.
The text is not inserted
- Check that the cursor is in a text field.
- Some apps block pasting into certain fields, such as password fields. This is intended.
- If no text field is active at all, e.g. on the desktop, VoxaFlow puts the text on the clipboard and shows “No text field”. Paste it with
⌘V/Ctrl+V.
Mac
- In Screen Sharing to another Mac, VoxaFlow types the text with this Mac’s keyboard layout. The other Mac needs the same layout (for example both U.S.), otherwise characters like “@” or “z” arrive wrong. The shared clipboard is not needed for this.
- The last text is available in the menu under “Paste last dictation again” and “Copy last dictation” and in the history.
Windows
- If the program runs as administrator, VoxaFlow cannot insert text there. The text is then on the clipboard; paste it with
Ctrl+V. - In Remote Desktop and Citrix, VoxaFlow types the text into the session. You set how under Packages › Remote Desktop and Citrix.
- The last text is available in the menu of the icon in the notification area under “Paste last dictation again” and “Copy last dictation” and in the history.
VoxaFlow isn’t running after a restart
Mac
- Turn on “Open VoxaFlow at login” under Settings › General.
- If it says macOS still has to allow this, click “Open Login Items” and switch VoxaFlow on under System Settings › General › Login Items.
Windows
- Turn on “Open VoxaFlow at sign-in” under Settings › General.
- If VoxaFlow is still missing, check in Task Manager under “Startup apps” that VoxaFlow is enabled there.
The text appears in the wrong language
VoxaFlow detects the language of your dictation and keeps it during cleanup. With very short sentences the detection can be wrong. In that case set the dictation language to English or German under Settings › AI.
English words in the middle of German text
When you speak German quickly with many English technical terms, the speech recognition sometimes writes similar-sounding English words, such as “by” for “bei” or “so that” for “sodass”.
The AI cleanup brings such words back to German; technical terms, names and code stay as you say them. This works best with VoxaFlow Lektor (Settings › AI).
Speak these passages a little more slowly. Add names that often come out wrong to the dictionary, with their typical mistakes.
The start sound is missing or annoying
Under Settings › Audio › Sounds you can choose a different start sound or “No start sound”. Also check there that the sound output points to the device you are listening on.
A word is always misrecognized
Add it to the dictionary, with the typical mistakes under “Misheard as”.
Cleanup takes a long time
On first launch VoxaFlow downloads the models, which can take a few minutes. The menu shows the progress.
Mac
After that, short dictations usually take less than a second.
Check the selected model under Settings › AI. Apple Intelligence was considerably slower in tests. VoxaFlow Lektor Kompakt needs about half the time of VoxaFlow Lektor but is slightly weaker with specialist terms; on Macs with 8 GB of memory it is the better choice. In Low Power Mode cleanup takes about twice as long.
Windows
After the installation VoxaFlow also spends a few minutes preparing the texts for the language model. During this time the first dictations take longer.
Check the selected model under Settings › AI. VoxaFlow Lektor Kompakt is faster than VoxaFlow Lektor but slightly weaker with specialist terms; on PCs with little memory or an older processor it is the better choice. In energy saver mode and on battery cleanup takes longer; check under Settings › AI that “Full performance on battery while dictating” is on. Close programs that currently need a lot of computing power.
Send feedback
Mac
If you get stuck or have a wish, choose “Send Feedback …”: at the bottom of the sidebar of the VoxaFlow window or in the menu of the menu bar icon. Pick “Problem”, “Idea or wish” or “Other”, briefly describe what you did and what happened, and click Send. If you enter an email address, we can write back.
For a problem VoxaFlow attaches a diagnostic report: version, macOS version, Mac model, settings and the log entries of the last 24 hours. Dictated text, audio, history, dictionary and snippets are not included, only their number. The names of your audio devices are included because they help with microphone problems. “View …” shows the report first; without the checkbox it stays on your Mac. Nothing is sent until you click Send. If the server cannot be reached, “Send by email instead …” sends the same to hello@voxaflow.ai.
Windows
If you get stuck or have a wish, choose Help › “Send Feedback …” in the menu of the icon in the notification area. Pick “Problem”, “Idea or wish” or “Other”, briefly describe what you did and what happened, and click Send. If you enter an email address, we can write back.
With the checkbox “Attach diagnostic report” a report is included: version, Windows version, processor, memory, settings and the log entries of the last 24 hours. Dictated text, audio and history are not included. “View …” shows the report first; without the checkbox it stays on your PC. Nothing is sent until you click Send. If the server cannot be reached, “Send by email instead …” sends the same to hello@voxaflow.ai.