You've been explaining the settlement logic for twenty minutes straight, and it's clicking. Then you hit 1500 words and the dictation tool cuts off. You switch back to typing, and the momentum breaks. You finish the doc at 2am, and it shows.
This is what's changed for developers in the past two years. Code used to be the hard part. Now it's the specs. The prompts. The design docs. The intent. You're not typing code anymore. You're typing 800-word explanations of what the model should build.
Typing those explanations is slower than speaking them. So you reach for a voice tool. Then you meet the paywall.
The word cap is a business model, not a technical limit
Wispr charges $14 a month for unlimited transcription. Willow charges $12. Superwhisper is $8.49 and still capped. The free tiers? 1500 words. 2000 words. Halved. None of them explain why, because it's not a technical reason. It's a pricing funnel.
Whisper, the open model those tools are built on, doesn't have a word limit. It's not metered. It doesn't cost per transcription. It runs locally on your device, which is why it's free. The company that makes Whisper could cap free users at 50 million words a day and it wouldn't cost them anything extra.
But caps are easier to monetize than trust.
What changed is the shape of your work
A year ago, developers complained about typing speed. That made sense when the work was translating design mocks into CSS or debugging a crash log. Type 60 words per minute, and voice-to-text bought you maybe 10 minutes per day.
Now your bottleneck is different. You spend 45 minutes on a design doc explaining why the settlement reconciliation needs eventual consistency, not strong consistency. You're not writing the code. You're explaining the constraints. The architecture. The trade-offs. The fallback paths if the payment processor is slow. That's 1200 words of intent. With a 1500-word free cap on Wispr, you get through the first two sections, then you're done.
With Recitey's local Whisper, you're not done. You keep going. The transcript is rough, but it polishes into clean prose in under 2 seconds. You hit send.
The technical credibility is the whole point
You're skeptical of voice tools because most of them handle your code IP like any other SaaS product: it goes up to their servers. Processed in their privacy-compliant pipelines. Logged. Anonymized. And you don't know how anonymized, because you can't read the terms and they won't show you where the data actually goes.
Recitey runs Whisper locally on your device. Your words never leave your machine. No cloud transcription. No variable costs for the dictation layer. The only time your words go anywhere is when you ask for the paid rewrite, and that's optional. It's a choice, not a trap.
This is the piece that matters because you've got payment settlement code in your voice. You're not dictating it into a service that's optimized to get you on their retention treadmill.
The rewrite is the feature, not the dictation
This might sound backwards. Whisper is accurate, but it's Whisper. You still get the "ums" and the false starts and the sentences that branch three directions. The rewrite polish is what makes it usable.
That's the paid tier. But the free tier is the unbounded dictation. You're not paying to speak. You're paying to turn speech into publishable prose.
Most tools flip it backwards: they charge you to speak, then give you the raw transcript. Wispr charges $14 so you can use their free cap. Recitey charges you only for the rewrite. The dictation itself is free and uncapped because Whisper running locally is zero variable cost per utterance.
Why Cursor, not VS Code
You use Cursor instead of VS Code because tab-complete reduces the friction of explaining intent. You refuse to put code in cloud-based transcription. You want tools that understand the LLM era demands a different shape of work: more words, more intent, less typing.
Recitey was built for the same reason: local-first transcription, uncapped, no cloud IP risk.