Developers used to write code on keyboards. Now they write intent. The work shifted, but most voice tools didn't: they still charge by the word, cap the free tier, or run in the cloud as if transcription cost were elastic. The bottleneck is no longer typing speed. The bottleneck is now the 1200-word design doc about payment settlement edge cases, the one you abandoned at 11pm because your tool's metrics system got in the way.
The design doc that never got written
Marcus, a backend engineer on payment systems, knows this frustration intimately. His 11pm design docs are where the thinking happens. Architectural trade-offs, error-handling edge cases, how to handle concurrency in the rebalancing loop, these get clearest when he can speak them aloud while pacing. Voice is faster than typing when the idea is still forming.
But here's the thing: Marcus refuses to use cloud-based transcription. Code IP leaving the device is a non-starter with security teams. Superwhisper ($8.49) runs locally, but it's indie and doesn't integrate well with his workflow. Willow ($12) runs cloud-only. Most of the uncapped free options? Cloud, or capped, or both. So Marcus didn't use voice for design docs at all. He typed them, lost the momentum that voice gives, and carried half-formed thinking into the next morning.
Local-first is not a luxury, it's a compliance boundary
Developers know this: if your data leaves the device to get transcribed, then your code IP is following it. Even if the company pinky-promises to delete it, you now have a compliance conversation with your security team. Some teams just say no. The transcription model doesn't change this calculus. Whether it's Whisper, OpenAI's model, or anything else, if the inference runs in the cloud, your design doc lives in someone else's system for a moment, and that moment is enough.
Whisper (the model that runs inside Recitey's free tier) gets about 96.3% word accuracy on LibriSpeech. It's also the model most developers already trust, because it has been open-source for years and every indie speech tool worth using is built on it. Local execution means you're not paying a cloud fee for a model you could already run on your laptop. And you're not explaining to security why a half-written design doc needs to be transcribed in the cloud.
The free tier is the business model, not the loss leader
Most SaaS companies price their free tier as a taste of paywalled features waiting behind the pricing wall. But Recitey's free tier is the transcription itself, Whisper, local, uncapped. No word counter. No monthly ceiling. You can draft 200 design docs for free if you want to. The paid tier is not the dictation. The paid tier is the cloud rewrite, the cleanup that turns rough voice into polished prose. It's optional.
Marcus can now draft design docs by voice at 11pm without worrying about data leaving his device, without hitting a meter, without having to explain anything to his security team. The thinking flow stays intact.
The cursor problem
Marcus uses Cursor instead of VS Code specifically because Cursor's tab-complete reduces how much he has to rewrite after voice dictation. That choice should matter. A voice tool that only works in a browser, or only in one IDE, is designing for the old way, where voice was a side channel. Cursor is where intent goes now. So is Linear, GitHub, Slack, Notion. A tool that meets you in all of them is designing for the new way.
Why no metering
Cloud transcription tools meter because cloud transcription has variable costs. Every API call to a transcription service costs money. But Whisper runs locally. Zero API calls. Zero variable cost. So Recitey's free tier can afford to have no cap, because there's no reason to cap it. The economics are structural, not a pricing strategy.
When you hit a word limit on a dictation tool, the message isn't "this is a feature boundary." The message is "we're metering transcription cost down to you because we're paying for cloud." An unlimited free tier is a signal that the model runs on your device, the cost structure is different, and the company is being honest about the economics rather than hiding it behind a paywall.