You're explaining a payment settlement algorithm to a colleague, it's 11pm, the thinking is flowing, and you're getting the architecture down faster by voice than you ever could by typing. Then your dictation tool cuts you off mid-sentence. Free tier. Word limit. You're back to typing again.
The Bottleneck Shifted
The work of being a developer changed. Five years ago, the constraint was typing speed. Now it's clarity of intent. You're writing longer: design docs, PR descriptions, prompt engineering for Copilot, detailed bug investigation notes for Slack, postmortems. The keyboard is no longer the bottleneck. It's the cognitive load of holding the whole thought while your fingers try to keep up.
Voice is faster for this. Not because talking is inherently faster than typing, but because explaining something out loud forces you to think linearly. You can't backspace an idea halfway through. You have to finish the thought.
The problem is that most dictation tools, Wispr ($14 pro tier required for unlimited), Willow, Superwhisper, are priced like the bottleneck is still typing speed. They put a word cap on the free tier. 2,000 words a month. 1,500 if you're unlucky. You hit the limit halfway through a design doc.
What Local Actually Means
Recitey runs Whisper directly on your device. No API calls, no metering, no server checking a word counter. The large model that would cost money to run in the cloud costs you zero per additional word. The variable cost per dictation is literally zero.
This is why most cloud tools cap you. Every transcription costs them money: compute, bandwidth, storage, model inference. Wispr needs to charge because their architecture is cloud-first. They offer uncapped tiers, but only at $180 per year. That's the unlock, not the norm.
Recitey's free tier has no word limit because the cost structure doesn't require one. The pro feature isn't transcription. It's the rewrite, the cleanup pass that polishes rough voice into clean prose. That's the cloud part. That's what costs money to run at scale.
The distinction matters. Most developers I've talked to feel uncomfortable sending raw code or architecture thinking through cloud dictation. Not paranoia. Rational caution. You own the code. Your company owns the thinking. Why send both through someone else's servers?
The Workflow Shape
Marcus, the backend engineer doing payment settlement work at a Series B fintech in Stockholm, uses Cursor instead of VS Code specifically because Cursor's tab-complete reduces the number of rewrites he has to do after voice input. He sketches ideas by voice, Cursor predicts what he's trying to build, and his voice stays on track.
He used Superwhisper for a while. Hit the cap three times in a single design doc. Switched to typing. Lost the momentum of thinking out loud.
What he needed wasn't a higher tier. He needed no limit at all. And a tool that didn't send every word through an external server. Both matter for his workflow.
The Skepticism Deserved
Developers are allergic to tools that hide how they work. If your dictation tool won't show you the model it uses, or it keeps the prompt locked, or it sends everything to the cloud by default, developers will leave.
Whisper is open. The model weights are published. Anyone can run it locally. There's no mystery. There's no hidden rewriting happening server-side. What you say gets transcribed by a model you can inspect.
That transparency matters. It's not paranoia. It's the difference between a tool you trust and a tool you hope is trustworthy.
The Trade-off
Free doesn't mean "as good as pro." The rough transcription is usually good, but it catches accents inconsistently, sometimes hallucinates punctuation, and occasionally misses context that a human would catch instantly. The local model is 96.3% word-accurate on standard English, which is high, but you might be the 3.7% on a given word.
That's why the rewrite exists. That's why the paid tier exists. It's the editorial pass. You dictate messy, the cloud makes it clean. Free is the capture. Pro is the polish.
For a design doc that needs to be perfect, or a PR description that hundreds of engineers will read, the rewrite is worth five or ten dollars a month. For your quick Slack explanations of why the migration failed at 2am, or your note-to-self about an edge case, uncapped free is everything.
The real shift in developer work wasn't that we should all talk faster. It was that we need to think longer. A tool that respects that, that doesn't interrupt you mid-thought with a word counter, stops being a productivity app and starts being part of how you work.