Vokalu vs Wispr Flow: memory and CPU on Mac and Windows

Vokalu used less memory and CPU than Wispr Flow in every scenario in our latest set of resource tests, using the Windows 0.2.21 rerun for the window-open scenario. It also occupied less space on disk. On the M3 MacBook Air, background memory averaged 110 MB versus 999 MB. On Windows, the updated app’s open-window CPU median was 1.26% versus 2.67%.

We make Vokalu and ran these tests ourselves. They describe particular versions on two computers, under the workloads described below. The Mac runs used Vokalu 0.2.19; the Windows background and dictation runs used 0.2.20, with the window-open test repeated on 0.2.21. Wispr Flow was 1.6.1034 throughout. The scope is memory, CPU and installed size during background use, open-window use and dictation.

Mac: background, open window and dictation

Tested October 2, 2026, on a MacBook Air (Mac15,12), Apple M3, 8 GB RAM, macOS 27.0.1. Vokalu 0.2.19 was the universal build running natively; Wispr Flow 1.6.1034 was the Apple silicon build. Both were signed in, had Microphone and Accessibility permissions, had dictated before testing and had run for at least six minutes.

Memory is the average physical footprint, summed across each app’s processes and its attributed helpers. CPU is a percentage of one core; idle rows use medians and dictation rows use averages.

Mac resource measurements, October 2026

Scenario	Vokalu memory	Flow memory	Vokalu CPU	Flow CPU
Background, windows closed, 5 min	110 MB	999 MB	1.47% median	6.13% median
Main window open, 2 min	139 MB	942 MB	0.15% median	6.30% median
Dictation, windows open, 6 min	191 MB	992 MB	5.5% average	13.2% average
Dictation, windows closed, 6 min	145 MB	967 MB	5.7% average	12.2% average

Background memory was about 89% lower in this run: (999 − 110) ÷ 999. During dictation with the windows closed, it was about 85% lower. Memory peaked at 193 MB for Vokalu and 1,140 MB for Wispr Flow in that dictation run. App sizes on disk were 52 MB and 511 MB respectively.

Vokalu still used about 1.5% median CPU in the background on this Mac. That is an area we want to improve; a lower result than another app does not mean zero overhead.

Windows: the 0.2.21 update and rerun

Tested October 2–3, 2026, on an Intel Core i7-14700F desktop with 28 logical cores, 32 GB RAM and Windows 11 Home, build 26200. Both apps were signed in and had dictated before testing.

Memory is the average private working set, including child processes. This differs from the Mac metric: compare apps within each platform, not Windows numbers against Mac numbers. CPU again means a percentage of one core, not of the whole 28-thread machine.

Windows resource measurements, October 2026

Scenario	Vokalu version	Vokalu memory	Flow memory	Vokalu CPU	Flow CPU
Background, windows closed, 5 min	0.2.20	119 MB	569 MB	0.9% median	2.1% median
Main window open, 2 min, latest rerun	0.2.21	172 MB	433 MB	1.26% median	2.67% median
Dictation, windows closed, 6 min	0.2.20	131 MB	619 MB	3.0% average	6.6% average

In the 0.2.21 window-open rerun, Vokalu used 53% less median CPU and about 60% less memory. Background memory in the 0.2.20 run was about 79% lower. The install folders measured 24 MB for Vokalu and 584 MB for Wispr Flow.

The first window-open run on Vokalu 0.2.20 was worse: 8.0% average CPU versus 2.6%, or 7.5% versus 2.4% using medians. We traced that to the continuously animated streak flame on Home. Version 0.2.21 limits the animation, and the rerun measured 1.5% average versus 3.2% (the 1.26% versus 2.67% medians above). These are two distinct runs; Wispr Flow’s memory changed too. We retain both in the downloadable measurements.

How we measured

Each measurement run sampled both running apps. The Mac script summed physical footprint for the app, its helpers and WebKit helpers attributed to it. The Windows script summed private working set for the app and its child processes, including the embedded browser helpers.

Mac sampling was once per second, or every half-second during dictation. Windows sampling was about every 2.3 seconds for background use, 1.4 seconds with the window open and 0.8 seconds while dictating after improving the collector. CPU figures are per core, so a process using multiple cores can exceed 100%.

Each six-minute dictation run included five messages per app, Vokalu first. The averages cover the entire interval, including pauses, rather than only active speech. We used these same scripts for both apps:

1. “Hey, quick update. The design review moved to Thursday at two, so let’s push the launch checklist to Friday morning.”
2. “Can you send me the latest numbers from last week? Um, I mostly need the sign-ups and how many people actually stayed.”
3. “Let’s meet at the coffee place on Main Street at ten. Actually, make it ten thirty, I’ve got a call that might run over.”
4. “Reminder to self: renew the domain, update the pricing page, and reply to Sarah about the podcast interview.”
5. “I think the new onboarding feels a lot smoother. The permissions step is clear, and the install window finally looks premium.”

Inspect the Mac samples and collector or the Windows samples and collector. These public files contain numeric resource samples and the measurement scripts, not dictated text or account data.

Reading the resource measurements

We used one computer per platform and one run per scenario, plus the Windows 0.2.21 rerun. Both apps were running, with Vokalu dictated into first. The tables describe those recorded sessions; background activity, app state, hardware and future releases can affect resource use.

Disk sizes also need context. Vokalu uses the operating system’s web view; the Windows install-folder figure excludes the shared WebView2 runtime. Wispr Flow bundles its browser runtime. Optional Vokalu speech models are separate downloads and are not part of these install-folder figures. Installed size is measured separately from optional on-device speech models.

For an app that stays open while you work, memory, CPU and disk usage are practical parts of the experience. In these sessions, Vokalu had the smaller resource footprint. For features and pricing, see Vokalu vs Wispr Flow. For installation and current Mac differences, see Vokalu for Mac.