Home / Support / AI Modes

AI Modes

Overview

Keyword = Access is triggered when you mention that word in your prompt (e.g., "Answer what's in my clipboard" or "Answer based on selection").

Write Mode

Rewrite Mode

Answer Mode

Reply Mode

Run Command Mode

Custom Modes

You can create your own AI modes — and fully edit the built-in ones — in Settings → AI Modes. Click Add Mode to create one, or Edit on any mode to change it.

What you can configure

• Name and Enabled — what the mode is called, and whether it's active.
• Trigger words — the word(s) you say at the start of your dictation to activate the mode (e.g. "translate", "summarize"). Say the trigger, then your instruction.
• Model — the included Qwen 3.6 model (free and fast), or any model from a provider you've connected (see below).
• Prompt — the instructions sent to the AI. Use the [CONTEXT] token to mark where your clipboard or selected text is inserted. If a mode includes context but you don't place the token, it's added at the end automatically.
• Include context — for Clipboard and Selection, choose Never, Always, or When speech contains a specific word.
• Output — Type it (virtual keyboard), Show in the WhisperTyping window, or Run as a command.

Bring your own AI provider

By default, modes use WhisperTyping's included Qwen 3.6 model, with responses capped at around 1,000 words. To use other models, add your own API key under Settings → AI Modes → AI Providers:

Each provider page has a "get a key" link. Paste your key, click Test, then Save. Your key is stored encrypted on your PC, used to call the provider directly (their billing applies, and the ~1,000-word cap no longer applies), and is never uploaded to WhisperTyping. You can also enter a custom model id for any model a provider offers.

Run Command: how it works

When a mode's Output is set to Run as a command, the AI turns your dictation into a Windows command and runs it. The command has two parts:

• command — what to run. This can be a URL (opens in your default browser) or an executable — either an absolute path like C:\Tools\app.exe or just a program name like chrome.exe or notepad.exe.
• arguments — optional. The command-line arguments to pass to that command (left blank for a URL).

A few examples:

Building your own Run Command mode

If you create your own Run Command mode, your prompt must instruct the AI to reply with only a JSON object containing those two properties — command and arguments. Nothing else, no explanation. It should look exactly like this:

{
  "command": "https://youtube.com",
  "arguments": ""
}

A few more examples of the JSON the AI should return:

// open the calculator
{ "command": "calc.exe", "arguments": "" }

// open the WhisperTyping site in Chrome
{ "command": "chrome.exe", "arguments": "https://whispertyping.com" }

// open the downloads folder
{ "command": "explorer.exe", "arguments": "%USERPROFILE%\\Downloads" }

// run a custom program by full path, with arguments
{ "command": "C:\\Tools\\Backup\\backup.exe", "arguments": "--full" }

For a website, put the complete URL (including any query string) in command and leave arguments empty.

Because the result is executed on your machine, only use Run Command with prompts you trust.