Skip to content
Call wiki

What does Wispr Flow add over built-in dictation?

Last updated 2026-08-17 · Confidence: documented — Wispr’s help center and pricing page, Apple’s macOS user guide, and Anthropic’s help center; the ten-minute-dump workflow is Alejo’s first-hand practice.

Wispr Flow dictates into any text field on Mac or Windows and cleans up what you said — built-in dictation transcribes verbatim, and the Claude desktop app’s dictation only talks to Claude. Alejo recommends it: hold the hotkey, talk for up to 20 minutes, release.

How it works

Section titled “How it works”

Hold the hotkey (fn on Mac, Ctrl+Win on Windows, customizable; double-tap for hands-free) and speak; the text lands in whatever field has focus, in any app. After you stop, Flow cleans up filler words, punctuation, and self-corrections, and it learns your names and technical terms over time. It needs an internet connection — no offline mode.

vs macOS dictation

Section titled “vs macOS dictation”

Apple’s built-in dictation is free, runs on-device on Apple silicon with no length cap, and auto-inserts punctuation. But punctuation is the whole cleanup: fillers and self-corrections come through verbatim, and it stops after 30 seconds of silence.

vs Claude’s Caps-Lock dictation

Section titled “vs Claude’s Caps-Lock dictation”

The desktop app’s quick entry adds Caps-Lock dictation with real-time transcription (macOS 14+, off by default) — but it only composes messages to Claude. Flow types anywhere: Slack, Docs, a terminal. More in Desktop vs browser.

Limits and pricing

Section titled “Limits and pricing”
  • Sessions run up to 20 minutes on desktop (warning at 19; at the cap Flow transcribes and pastes what it has), on free and Pro alike.
  • Free plan: 2,000 words/week on desktop. Pro: $15/month, or $12/month billed annually, unlimited.

Alejo’s recommendation

Section titled “Alejo’s recommendation”

Alejo recommends it for long thought dumps: when starting something messy, hit the hotkey and talk for ten minutes, then let an agent structure the result. Speaking also beats typing for agent prompts — people give more context out loud, and context is what the model needs.

Sources

Section titled “Sources”