Browser dictation: when speaking beats typing (use cases)

November 10, 2026·by the QuickSpeak team·we make the dictation tool described here

You already talk faster than you type — roughly 2–3× faster for most people. The gap is mostly wasted on first drafts: the long email you compose at 50 words a minute, the support ticket written at the end of an exhausting call, the CRM notes you type while the details are still fresh. Dictation closes the gap by putting your speech straight into the text field — and when it runs on your machine, it does so without sending your words anywhere. Here's where it genuinely pays, and where it doesn't.

The short version

  • QuickSpeak's dictation: hotkey on, talk, hotkey off — text inserts at your cursor in any input, textarea, or rich-text editor on any site. Same on-device model as the meeting transcription; nothing uploads.
  • Best targets: long emails, support-ticket replies, CRM notes right after calls, message threads, first drafts of documentation.
  • Still type: code, tables, spreadsheets, anything you're editing word-by-word rather than drafting.
  • Dictation is a Pro feature ($12/mo or $99/yr, 14-day trial no card); meeting transcription stays free without limit.

The use cases that actually pay

  1. The end-of-call email. Right after a client call, the summary email takes eight minutes typed and three dictated — and the dictated version is better, because the call is fresh and speaking pulls detail out of memory that typing filters away. Same logic as the transcript-to-minutes workflow, but for calls you didn't record.
  2. CRM and ticket notes at volume. Sales and support roles live or die by notes-in-the-system, and notes are exactly the kind of prose nobody wants to type. Dictating into Salesforce, Zendesk, or HubSpot right after each interaction turns "I'll batch the notes Friday" into done-now — and Friday-you finds real sentences instead of fragments.
  3. Long message threads. Slack and Teams discussions where you're writing paragraphs, not one-liners. Voice handles the paragraph; typing handles the "sounds good 👍".
  4. Documentation first drafts. The README, the runbook, the process doc — the hardest part is the blank page. Speaking a rough draft and then editing at the keyboard is faster and usually better-structured than typing the draft from scratch. (Editing stays a keyboard job — see the honest limits below.)
  5. Accessibility and strain. For colleagues with RSI, wrist strain, or temporary injuries, dictation isn't a speed tool — it's the way work continues. Same spirit as captions as an aid: one tool serving several needs, with the person using it deciding what works.

How QuickSpeak's dictation works, mechanically

A hotkey toggles dictation on and off (the toggle always stops first, even mid-processing). Whatever you speak is recognized by the same Parakeet model that powers the meeting transcription — running locally in WebAssembly — and the text inserts at your cursor in the focused field: inputs, textareas, and rich-text editors across any website. The insertion uses caret-accurate placement with native undo, so a misfire is one ⌘Z, and React-style controlled editors stay consistent. Because recognition is local, dictation works on a plane and — more importantly for work tools — nothing you dictate ever reaches a third-party server. That last property matters when the text field belongs to an HR system, a legal tool, or anything medical-adjacent: check what a dictation tool does with your words before using it there. Cloud dictation services send every sentence to someone's server; local dictation doesn't have that question to answer.

Where typing still wins, honestly

  • Code and anything symbolic. Dictating TypeScript is a party trick, not a workflow. Symbols, indentation, and references don't survive speech well.
  • Tables, spreadsheets, forms. Field-hopping is faster with the keyboard; dictation shines on prose-in-one-place.
  • Precision editing. Replacing one word, reordering a clause, nudging a number — the keyboard does this in one motion; voice needs several. Draft by voice, edit by hand.
  • Open offices. Dictation is audible. Headphones help; a shared desk at 9am does not.

One workflow note: dictated drafts arrive as long breath-driven sentences. Plan on a light punctuation pass — the accuracy failure modes are the same ones as any speech recognition, mostly names and terms, which the free custom-vocabulary list fixes across both dictation and meeting transcription.

FAQ

How does browser dictation work in QuickSpeak?

Hotkey on, talk, hotkey off — text inserts at your cursor in whatever field is focused (inputs, textareas, rich-text editors), on any site. Recognition runs on your machine. It's a Pro feature ($12/month or $99/year, 14-day trial no card).

What is dictation faster than typing for?

First drafts and one-way prose: emails, ticket replies, CRM notes, messages, documentation drafts. Speaking is roughly 2–3× faster than typing for getting words out. Code, tables, and fine editing stay keyboard jobs.

Is dictating into work tools private?

With on-device dictation, yes — nothing leaves your machine. Cloud dictation tools send every sentence to a third party; check that before dictating into HR, legal, or medical-adjacent systems.

Does it work in any web app?

Standard inputs, textareas, and contenteditable editors — which covers Gmail, Zendesk, Notion, Salesforce, Slack, and most web apps. Exotic custom editors may fall back to simpler insertion.

Sources & notes

  1. QuickSpeak dictation behavior (hotkey toggle, cursor-accurate insertion in inputs/textareas/contenteditable, on-device recognition, Pro entitlement): the shipping extension — `src/dictation/` and `src/license/types.ts`, accessed November 10, 2026.
  2. Speaking-vs-typing speed ranges (roughly 120–150 wpm speech vs 40–60 wpm typing) are widely cited general figures used here for ordering, not precision.