← Back to VoicePaste

Help

Everything you need to get started with VoicePaste, on macOS and Windows.

Showing the macOS instructions. On Windows? Switch above. Showing the Windows instructions. On a Mac? Switch above.

Getting Started

  1. Download and install VoicePaste. Open the downloaded DMG and drag VoicePaste to your Applications folder.
  2. Launch VoicePaste. On first launch, macOS may ask you to confirm since the app was downloaded from the internet. Click Open. VoicePaste is notarized by Apple, which means Apple has scanned it for malicious software before distribution. VoicePaste lives in your menu bar (top-right of the screen).
  3. Grant permissions. VoicePaste asks for Microphone and Accessibility access. Both are needed for full functionality. See Permissions below.
  4. Press your hotkey and speak. During setup VoicePaste suggests V with auto-detect language. Press it to start recording, speak, then press it again to stop. Your text appears wherever your cursor is.
  1. Download and install VoicePaste. Run the downloaded .exe installer. The first Windows builds are not signed yet, so Windows SmartScreen may warn you: click More info, then Run anyway. It installs for your user account, so no administrator rights are needed.
  2. Launch VoicePaste. Open it from the Start menu. VoicePaste lives in your system tray (the notification area at the bottom-right, next to the clock). Click the arrow there if you don't see the icon.
  3. Grant microphone access. Windows only needs the Microphone permission, no Accessibility step. See Permissions below.
  4. Press your hotkey and speak. During setup VoicePaste suggests CtrlShiftV with auto-detect language. Press it to start recording, speak, then press it again to stop. Your text appears wherever your cursor is.
VoicePaste works out of the box with Cloud mode. No API keys or account needed. You get 30 minutes of free transcription per month.

Permissions

Microphone

Required to record your voice. Without it, VoicePaste cannot capture audio.

macOS shows a system dialog the first time you try to record. If you miss it, grant access in System Settings → Privacy & Security → Microphone.

Enable it in Windows Settings → Privacy & security → Microphone, and make sure "Let desktop apps access your microphone" is turned on.

Accessibility

Required for auto-paste: typing the transcribed text directly into your focused application. Without this permission, VoicePaste will still copy text to your clipboard, but you'll need to manually paste with V.

To grant Accessibility access: open System Settings → Privacy & Security → Accessibility, and enable VoicePaste.

VoicePaste is distributed outside the Mac App Store specifically because the App Store's sandbox blocks the Accessibility API needed for auto-paste. This is the same reason apps like Raycast, Alfred, and BetterTouchTool distribute directly.

No Accessibility permission needed

On Windows, VoicePaste types the transcribed text into your focused window directly. There is no extra permission to grant and no administrator rights required. Microphone access is the only permission VoicePaste needs.

Hotkeys

VoicePaste supports up to 10 global hotkeys. Each hotkey defines:

For example, you could have:

Configure hotkeys in Settings → Hotkeys & Writing. You can add, remove, and reorder them freely.

Quiet other audio while recording

If you dictate while music or a video is playing, the microphone can pick it up and muddy the transcription. VoicePaste can automatically quiet other audio while you record, then bring it back when you stop.

Open Settings → General → Recording and choose what happens to Other audio while recording:

Lower volume turns down your computer's overall output volume, so it quiets everything at once: music, videos, browser tabs, games, and system sounds. Whatever you pick, VoicePaste always restores your audio afterward. Stopping, cancelling, an error, or quitting mid-recording all bring it back.

The first time you use Pause, macOS shows an Automation prompt asking to control your media app (for example, "VoicePaste wants to control Spotify"). Click OK to allow it. If Pause does nothing, check System Settings → Privacy & Security → Automation. The Lower volume option does not need this.
Pause uses the standard Windows media controls to pause whatever is playing, so there is no extra permission to grant.

Writing Styles & Translation

Each hotkey can optionally restyle the transcription, translate it, or both, in a single press. The original spoken text is always preserved in History so you can undo or re-style afterward.

The six writing styles

Pick None to keep the raw transcription unchanged.

Example. You speak "hey just wanted to follow up on the deck I think we agreed I'd send it monday morning". The Professional style turns that into "Following up on the deck. As agreed, I'll send it Monday morning."

Translate as part of the hotkey

A hotkey can also translate the result into a target language before it is pasted. The pipeline is:

  1. Speak (in any supported source language)
  2. Transcribe
  3. Restyle (if a style is selected)
  4. Translate (if a target language is selected)
  5. Paste

Style and translate can be combined. The style is applied in the target language, not the source. So "Auto-detect → Spanish, Professional" on Dutch input gives you a professional Spanish output.

Re-style or re-translate later

If you pick the wrong style or language at recording time, open History or press your Transcription Clipboard hotkey, pick the entry, and re-style or re-translate from there. The original text is kept so re-runs always start from the source, never from a previous translation.

Configure in Settings → Hotkeys & Writing, then pick a hotkey and set Writing Style and Target Language.

Speech Cleanup

Natural speech is full of "um", false starts, and mid-sentence corrections. Speech Cleanup tidies the raw transcription without changing what you meant. It runs in Cloud mode, before any writing style or translation, and it never summarizes or rephrases your content.

Turn it on with the Clean up speech switch in Settings → Transcription, then choose which passes to apply:

Speech Cleanup only cleans up delivery. To change tone or wording use a Writing Style; to force exact replacements use Text Replacements. The original spoken text is always kept in History, so nothing is lost.

Speech Cleanup is a Cloud feature. In Local Whisper and own-API-key modes the transcription is returned as spoken, so use a Writing Style or Text Replacements there instead.

Dictionary

The Dictionary lets you teach VoicePaste names, brands, technical terms, and product names that transcription engines tend to mishear. Add a word once and it stops coming back as something close-but-wrong.

How to add entries

Open Settings → Transcription and type the words you want recognized, one per line or separated by commas. Entries are matched case-insensitively, deduplicated, and trimmed. The exact casing you type is what VoicePaste hands to the transcription engine, so use the spelling you want pasted (for example, iPhone not iphone, OpenAI not openai).

How it actually works

Different transcription engines use the dictionary differently. VoicePaste handles each one for you. The behaviour differs:

What the dictionary cannot fix

If a word is consistently misrecognized in a way the dictionary doesn't catch (different word, different intent), use Text Replacements instead. The Dictionary nudges the engine; Text Replacements is a hard find-and-replace after the fact.

Text Replacements

Text Replacements are find-and-replace rules applied to every transcription before it is pasted. Use them for things the Dictionary can't fix: words that the engine recognizes correctly but consistently produces in a form you don't want.

Setting them up

Open Settings → Text Replacements. Each rule is a pair:

Example. The transcription keeps coming back as "Bitaccess" when you say "BitAxes". Add a rule: Find Bitaccess → Replace with BitAxes. From then on every transcription with Bitaccess is corrected automatically.

How matching works

When it runs

Text Replacements are applied after transcription and after rewriting / translation, immediately before the text is pasted. They also run on the text shown in History and the Transcription Clipboard, so re-pasting an old entry uses your current ruleset.

Transcription Clipboard

Press K from anywhere on your Mac to open a floating window with your recent transcriptions. It's the fastest way to recall something you said earlier, translate it into another language, or fix a mistake without re-recording.

Press CtrlShiftK from anywhere in Windows to open a floating window with your recent transcriptions. It's the fastest way to recall something you said earlier, translate it into another language, or fix a mistake without re-recording.

Inside the popup

Change the popup hotkey in Settings → General if it conflicts with another app.

The clipboard is fully local. Your transcription text and audio stay on your computer. Even when translation or restyling runs through VoicePaste Cloud, the result is only stored back into your local history.

History & Export

Every transcription is saved to your local History, with its text, language, writing style, and the original audio. History lives only on your computer; nothing is uploaded.

Finding past transcriptions

Control whether history is kept, and for how long, with the Save transcription history and retention options in Settings → History.

Export

Export your history to a file from Settings → History. Choose how much detail to include:

Transcribe recordings you already have

VoicePaste can transcribe existing audio, not just live dictation:

Using VoicePaste with AI

This is the use case VoicePaste is best at. If you spend hours a day talking to AI tools, voice is the fastest connection you have. AI is excellent at giving structure to thoughts that don't yet have any. You no longer need to formalize what you want to say before you say it: dump first, let the model order it.

Where VoicePaste pastes into

Anywhere your cursor blinks. That includes:

Tools like ChatGPT desktop ship with their own voice input. That works until you need to send the result somewhere else and end up copy-pasting between windows. VoicePaste skips the middle step.

The brain-dump workflow

Keep a scratch text file open (or just an empty AI chat). Talk into it as you think, test, or debug. Don't worry about ordering, completeness, or grammar. When you're done, ask the AI to extract action items, summarize, or structure the dump into a document.

This works because most of the cost of writing isn't typing the words. It's deciding what to write. Voice removes the typing cost. AI removes the structuring cost. What's left is your raw thinking.

The multi-window workflow

If you run several AI sessions in parallel, switching between them is just standard focus switching: TabAltTab, a click, or your window manager. The VoicePaste hotkey stays the same regardless. Focus the window you want, press your hotkey, talk, the words go into whichever window has focus. One hotkey, muscle memory, any number of AI sessions.

Interrupted? Recover from history.

Mid-thought interruptions are common when you work across many AI windows. Press Esc to cancel a recording. Nothing is lost: open the Transcription Clipboard and pick up where you left off. You can favorite (F) the entries you want to keep around as references.

What works well to dictate to AI

Voice is high-bandwidth, but the deeper insight is that it lets you self-correct in flow. Typing forces you to assemble a structured sentence before you commit it. Voice lets you change direction mid-thought ("ignore the last bit, I meant…"), and the AI reconciles it. Use the keyboard for precision: editing code, picking exact words. Use voice for everything else.

VoicePaste Cloud vs VoicePaste Local

VoicePaste Cloud

The default. No API keys needed, no account to create. Audio is sent to the VoicePaste server for transcription. Usage is tracked with a monthly quota based on your subscription tier. 30 minutes per month for free.

VoicePaste Local

VoicePaste Local lets you transcribe without our servers:

VoicePaste Local works without any purchase. A small tagline is appended to each transcription: "Transcribed by VoicePaste." The $99 one-time purchase removes the tagline. Your history always stores clean text regardless.

Configure in Settings → VoicePaste Local.

Your Data and Privacy

Where your data lives

Your transcription text, audio recordings, history, settings, and API keys are stored locally on your computer. How audio is handled while transcribing depends on the mode you choose:

What VoicePaste Cloud stores

When you use Cloud mode, VoicePaste stores only the minimum metadata needed to run the subscription service:

VoicePaste Cloud does not store:

What about syncing across devices?

VoicePaste does not currently sync transcription history between devices. Because we don't store your transcriptions on our servers, there is nothing to sync from. Your history lives entirely on each computer.

This is a deliberate design choice for privacy. If we introduce syncing in the future, it will be built with end-to-end encryption so that only your devices can read your data.

Troubleshooting

Nothing happens when I press the hotkey

Text doesn't paste automatically

Transcription comes back in the wrong language

Transcription is inaccurate

VoicePaste isn't hearing me

Error: "No internet connection"

Cloud and own-API-keys modes need internet to transcribe: the audio is sent to the transcription provider over the network. If you're offline in either of those modes, recordings are saved locally and will be transcribed automatically when your connection returns.

Local Whisper mode is the exception. Once you've downloaded a model, transcription runs entirely on your computer and works fully offline. If you regularly dictate without internet, switch to Local Whisper in Settings → Transcription.

Licenses & Devices

Activating a license

After you buy, your license key is emailed to you. Open Settings and go to the account panel (VoicePaste Cloud for a subscription key, VoicePaste Local for a one-time key), paste the key, and click Activate.

Using VoicePaste on several computers

A paid license works on up to 5 devices. Each computer you activate is linked to your license. In the account settings you can see your linked devices, rename them, and remove ones you no longer use.

Moving to a new computer

Install VoicePaste on the new machine and activate it with the same key. If you have reached the device limit, remove an old device from the linked-devices list first, or sign out on the machine you no longer use, to free a slot.

The free tier and Local Whisper need no key at all. A license is only for paid Cloud plans and for removing the VoicePaste Local tagline.

Payments & Billing

How to pay

You can pay two ways. By card (and other methods) through Paddle, our merchant of record, who handles billing, invoices, and tax. Or with Bitcoin, paid on-chain directly, which is the cheapest option: it saves you about 25% versus card, with no card and no account needed.

Paying with Bitcoin

In the pricing section, switch the toggle to Bitcoin and choose a plan. Enter the email where your license key should go, then pay the on-chain amount: scan the QR code with any Bitcoin wallet, or copy the address. The page shows live status, and the moment your payment confirms on-chain, your license key is emailed to you. No account, no billing profile, just your email.

Bitcoin plans are a one-year prepayment (or a one-time purchase for VoicePaste Local). Bitcoin cannot auto-renew, so there is no subscription to manage: when the year is up, you simply pay again to extend, and we send a reminder before your access ends.

Renewals and cancellation

Card subscriptions renew automatically and can be cancelled anytime through the Paddle customer portal; you keep access until the end of the period you paid for. Bitcoin has nothing to cancel: it simply expires at the end of the prepaid year unless you choose to renew.

Refunds

Card payments are covered by our standard refund policy (for example, a 30-day money-back guarantee on VoicePaste Local). Bitcoin payments are final and non-refundable, because on-chain transactions cannot be reversed. So before buying with Bitcoin, try VoicePaste free first: Cloud gives you 30 minutes a month and VoicePaste Local is free, so you can confirm it works for you. If a Bitcoin order genuinely goes wrong, email support@voicepaste.app and we'll make it right.

Contact Support

Still need help? Email us at support@voicepaste.app and we'll get back to you as soon as possible.