I Tracked Wispr Flow’s Outages for Six Months. Here’s the Pattern.

A capable, well-liked dictation app with a structural reliability problem — and what it taught me about where voice software should run.

Why this matters: Wispr Flow logged 69+ outages in six months and holds a 2.7/5 Trustpilot score against a 4.8/5 App Store score. The gap isn’t bad luck — it’s architecture. Every dictation runs in the cloud, so one bad server day is everyone’s bad day.

Here’s a number that stuck with me: an independent uptime monitor, StatusGator, has logged more than 69 outages for Wispr Flow since it started watching on December 18, 2025 — roughly six months, with about 20 in the most recent 30-day window. The most-affected component is the one that matters most: dictation itself.

I want to be fair up front, because this is easy to get wrong. Wispr Flow is a genuinely good product. Lots of people love it. It scores 4.8/5 across 8,500+ ratings on the iOS App Store. But that same app sits at 2.7/5 on Trustpilot, and when a tool’s first-impression rating and its day-two rating disagree that sharply, the gap is the story.

The three signals I now use to judge any cloud dictation tool

After staring at six months of incident data, I landed on a simple lens. Don’t trust one headline number — weigh three signals together.

First, how often does the vendor’s own status page log incidents? This is the most honest signal, because the company is reporting on itself. Second, what’s the review trajectory, not the lifetime average? A 4.8-star lifetime score can hide a sharp drop in recent, organic reviews. Third — and this is the one people skip — does it stay good on day two? Reliability is a day-two property, not a day-one one.

Wispr Flow strains on all three. And the reason is the same each time.

A six-month cadence, not one bad week

What changed my mind wasn’t a single dramatic outage. It was the rhythm. A sign-in outage across all platforms in late March. Website and login errors in mid-April. Desktop startup failures at the end of April. A May 7 incident where Wispr’s own note named the cause: “Capacity degradation at an upstream provider.” Then a near-continuous latency run from late May into early June, with one day — June 2 — measuring roughly ten hours of degraded service.

That’s not one outage. It’s a six-month cadence of them, concentrated in the core dictation path.

To Wispr’s real credit, none of this is hidden. The company runs a transparent public status page, acknowledges incidents within about 15 minutes on average, and ships concrete fixes. I’d rather a vendor surface its problems than bury them. But transparent reporting tells you when dictation is down. It doesn’t keep dictation up.

Why the cloud inherits its server’s worst day

The recurring word in those incident notes — “capacity” — points at the root cause. When you dictate into a cloud tool, your audio leaves your device, hits a transcription pipeline, often passes through a large language model for formatting, and only then returns as text to your cursor. Every hop depends on a server having spare capacity at the exact moment you speak.

And because all users share that backend, one capacity problem degrades dictation in the US, Europe, and APAC at once. Restarting the app won’t help. Reinstalling won’t help. Switching Macs won’t help. The overloaded component isn’t on any of your devices.

This is the defining trade-off of cloud dictation, and it’s worth saying plainly rather than pretending it doesn’t exist. Cloud tools win on cross-platform reach and heavy AI rewriting. They lose on a thing you can’t patch from your end: a shared single point of failure.

The architecture I switched to — and why I build on it

This is the problem I started Voibe to solve, so treat this as the biased-but-honest part. Voibe is a Mac-native dictation app that runs OpenAI’s Whisper models entirely on your Apple Silicon Mac.

You press a hotkey, audio is captured into memory, transcribed locally, written into the active field, and discarded. There is no transcription server, because the model runs on your Neural Engine.

That changes the failure modes the article above describes. There’s no shared backend whose capacity can be exceeded, so the “regular outages” pattern has no equivalent. There’s no server-side processing, so there’s no mechanism to throttle you after the trial. It works on a plane, in a dead zone, or during a competitor’s worst outage day — internet status is simply irrelevant to whether your words become text.

I’ll be honest about the trade-off, because it’s real: Voibe is Mac-only, and on-device dictation forgoes cloud-only conveniences like instant cross-device sync. If you need Windows, iOS, or Android, a cloud tool is still your option. Plenty of people keep one for reach and run an on-device tool for the work that can’t wait for a server to recover.

Should you stay on Wispr Flow?

My honest answer is: it depends entirely on how much downtime your work can absorb. If you dictate occasionally and your deadlines are flexible, Wispr Flow remains a capable cross-platform tool — just bookmark its status page so you can tell “Wispr is down” from “my network is down.” If you’ve already hit the day-two drop — accuracy slipping, the app “rewriting” what you say instead of transcribing it — that’s the signal the pattern is reaching you.

But if your dictation is something you depend on — for accessibility, for coding, for volume, for deadlines — the more dependable answer is an architecture with no server to go down.

That’s the case for on-device dictation, and that’s why Voibe is built that way: Whisper running on your Mac, fully offline, transcribing faithfully, $149 one-time. It cannot have the outages in this article, because there’s no cloud in the path. You can try it free here — no account, no card, no server in the loop.

For the full six-month timeline with every source linked, the original post on Voibe Resources has the receipts.