Typing meeting notes is slow (40 WPM) and steals attention from the conversation. Bot transcription is delayed and routes audio through someone else's cloud. There is a third option: dictate a summary directly into Notion, OneNote, or Obsidian, at 150 words per minute, locally on Windows.
Most people pick one of two bad options. There is a third that is faster and more private.
Average typing speed is 40 words per minute. You can capture a fraction of what is said and your eyes are on the keyboard, not the screen sharing the slide. Attention is split, the notes are sparse, and you miss the next point while finishing the last one.
A bot joins the call, captures everything, and produces a transcript later. Useful for archives but the summary lands hours after the meeting ended. Audio passes through a vendor's cloud, which is awkward for confidential discussions. Worth the trade for some teams.
Speak a one-line summary into your notes app right after each agenda item, at 150 words per minute. Text lands in Notion, OneNote, or Obsidian instantly. Audio never leaves your Windows PC. The fastest way to take meeting notes, if you have 20 seconds to spare.
Six reasons speaking your notes is faster than typing them
The average person speaks at around 150 words per minute and types at around 40. That is a 3.75x throughput advantage on every sentence. A 30-second meeting summary that takes 4 minutes to type takes 30 seconds to dictate.
While you dictate, your hands can drag the cursor to the right slide, click the next agenda item, or pull up a Slack DM. Typing locks both hands to the keyboard for the duration of every note.
You can keep watching the screen share while you speak. Typing pulls your gaze down to verify spelling, which is exactly when the next important point lands. Dictation keeps your attention where the meeting is.
Open your notes app once. Bind a hotkey. From then on, every meeting works the same way: press the hotkey, speak, release. No bot to invite, no link to copy, no service to log into.
StarWhisper processes audio on the Windows machine itself. No cloud upload, no transcript stored on a vendor's servers. Confidential meetings stay confidential by default.
Notion, OneNote, Obsidian, Apple Notes (via web), Roam, Logseq, Bear, Joplin, plain Notepad, anything that accepts a text input field. StarWhisper pastes wherever your cursor lives.
The average sustained typing speed for adult knowledge workers is 40 words per minute. That figure comes from typing-test aggregators that have measured hundreds of millions of attempts. Touch typists who have practiced for years can hit 60 to 80 WPM in a quiet room. Almost nobody hits that during a meeting, because part of your attention is on the speaker and another part is on whether you spelled the project name correctly.
Real meeting note typing is closer to 25 to 30 WPM, because you are constantly stopping to listen, glance at the slides, and check that what you wrote matches what was decided. That means a 30-second verbal point becomes either a one-line note that misses half the detail, or a three-minute typing exercise that causes you to miss the next two agenda items.
Voice dictation breaks that bottleneck. Spoken English averages 150 words per minute in conversation and up to 180 when summarizing. StarWhisper uses OpenAI Whisper running locally on your Windows machine, so spoken text becomes typed text in the same time it took you to say it. The result is that you can fully capture a one-minute discussion point in about 20 seconds of dictation, including a one-line decision summary and the rationale.
Tools like Otter.ai, Fireflies.ai, Fathom, and Microsoft Teams' built-in transcription all do the same thing: a bot joins your meeting, records every speaker, generates a verbatim transcript, and produces a structured summary. These are excellent products for what they do.
They are not, however, the same problem as "I need to capture my own thoughts about a meeting, fast." Bot transcription captures everything that was said. Personal dictation captures what you wanted to remember about it. The first is a record. The second is a note. You often want both, and they coexist nicely.
There is also a confidentiality dimension. Bot transcription means the meeting audio leaves the participants' devices, travels to a vendor's cloud, gets processed by AI models, and may be retained for some window. For sensitive meetings (HR conversations, board discussions, legal strategy, M&A), that is a hard sell. Personal voice dictation with a local engine sidesteps the issue entirely because nothing about the meeting audio is captured. You speak your own summary into your own machine, after the moment passes.
If your situation calls for a meeting-bot transcript, our guide to how to transcribe meetings covers the trade-offs and approach in detail. This page is about the other workflow.
Here is what fast meeting-note dictation looks like once you have StarWhisper installed.
The whole loop is under 30 seconds per note. A typical 30-minute meeting produces 6 to 10 useful summary lines, captured live, with zero context switching.
| Note length | Type (40 WPM) | Dictate (150 WPM) | Time saved |
|---|---|---|---|
| 10-word decision | 15 sec | 4 sec | 11 sec |
| 30-word summary | 45 sec | 12 sec | 33 sec |
| 60-word context note | 90 sec | 24 sec | 66 sec |
| 10 notes per meeting | ~10 min | ~2.5 min | ~7.5 min |
| 20 meetings per week | ~200 min | ~50 min | ~150 min |
Two and a half hours per week back in your calendar from a single workflow change. The numbers above assume a moderate note-taking style. Heavier note takers see proportionally larger gains.
Bot-in-meeting transcription tools like Otter, Fireflies, and Teams transcription are genuinely useful for use cases this workflow does not cover. If you need a verbatim transcript of who said what, a multi-speaker summary with action items extracted by AI, or a searchable archive of every meeting your team holds, those tools are purpose-built for that and they do it well.
StarWhisper is the right tool when you want to capture your own thinking, fast, with confidentiality. The two approaches solve different problems and they pair well. Many users run StarWhisper for personal notes and a bot transcription tool for the full-record archive on meetings where the team has agreed to it.
| Capability | Typing | Bot tools (Otter, Fireflies) | StarWhisper dictation |
|---|---|---|---|
| Speed of capture | ~40 WPM | Real-time but delayed delivery | ~150 WPM |
| Audio leaves your device | No | Yes | No (Local Mode) |
| Works on confidential calls | Yes | Risky | Yes |
| Lands in your notes app instantly | Yes | Post-meeting summary | Yes |
| Captures everyone else | No | Yes | No (only you) |
| Cost | Free | $10 to $30/mo | Free or $10/mo |
| Works offline | Yes | No | Yes |
StarWhisper writes text into whichever app has keyboard focus at the moment you trigger dictation. That means it works with every notes app on Windows that accepts a text input cursor. In practice, these are the ones meeting note takers use most:
HR professionals in particular run this workflow for back-to-back 1:1s, candidate debriefs, and policy meetings. See our notes for voice to text for HR managers for the specific patterns used in those contexts.
The cleanest pattern, when policy and team norms allow it, is to run both. The bot transcription provides the verbatim archive. The personal dictation provides the focused, decision-oriented notes that actually make it into the project tracker.
Practical division of labor:
After the meeting, the bot summary covers the discussion. Your StarWhisper notes cover what you are going to do about it. Together they replace the "I will write up notes later" task that almost nobody actually completes.
Because your meeting summaries often include sensitive context (compensation, performance, deal terms, personal matters), it matters where the audio ends up. StarWhisper runs Whisper locally on your Windows machine in Local Mode, which is the default. The audio you speak is captured, fed straight into the model, turned into text, and discarded. There is no upload, no log on a server, no transcript stored in a vendor's cloud.
This is structurally different from cloud transcription. With a cloud tool, your audio travels to a vendor's servers, gets processed by their pipeline, and may be retained for some period. For meetings where confidentiality matters (almost all of them, if you stop and think about it), local processing is the easier story to defend. We expand on this in the dedicated Whisper local vs cloud reference.
The free plan gives you 500 words per day, which is enough for most note takers to evaluate the workflow during a normal week of meetings. If you find yourself running into the limit, Pro is $10 per month or $80 per year, with no per-seat math and no upsell tiers. The full pricing detail lives on the pricing section of the homepage and on the Pro plan page.
There is no charge for personal evaluation, no credit card required to install, and no time limit on the free tier. The 7-day Pro trial unlocks unlimited dictation if you want to stress test a full week of meetings before committing.
StarWhisper runs on Windows 10 and Windows 11. The CPU-only path works on essentially any modern laptop. An NVIDIA GPU with CUDA support makes transcription faster but is not required. The Whisper model is bundled with the installer, so once you have downloaded the app you do not need to download anything else to use the local mode.
For older machines, the app picks the right model size automatically. If you want the technical detail of how local processing works and what hardware paths are supported, the privacy and offline features page covers it.
Full meeting transcription workflows for Windows, with and without a bot.
Fast notes for 1:1s, candidate debriefs, and confidential conversations.
Use StarWhisper alongside Zoom calls without disrupting the meeting.
Dictate directly into Notion pages, databases, and meeting templates.