Maria sounds sharp on video calls with her US enterprise clients. She's direct, articulate, confident. Then she moves to Slack, and something breaks. The follow-up message she drafts reads like a careful, diminished version of herself. She re-reads it four times before hitting send, each pass wondering if she sounds too formal, too cautious, too much like someone translating herself into English rather than thinking in it.
The Call vs. The Message
The friction isn't language proficiency. Maria speaks English fluently in real-time. She thinks on her feet, builds rapport, handles objections. Her last performance review named "exec presence on calls" as a particular strength.
But in Slack, something shifts. The written message demands precision she knows she has, but the medium doesn't feel like it belongs to her. It's not about correctness. It's about recognizing herself in what she writes. On a call, that happens naturally. In Slack, it takes work.
She reads the draft. It's grammatically sound. It's clear. But it doesn't sound like her. She rewrites.
She reads the second draft. Still careful. Still not quite right. She rewrites.
By the fourth pass, 30 minutes has gone by on a single message that took two minutes to explain in person.
What Grammarly Misses
Grammarly catches the comma. It flags the misplaced modifier. It polishes surface-level English and calls it done.
But Grammarly solves a problem Maria doesn't have. She isn't searching for the right word. She isn't uncertain about tense. Her performance review explicitly notes the gap: "could be more concise in writing." Not "more correct." More concise. As if the medium itself makes her verbose, defensive, over-explaining.
The real issue is voice. Grammarly removes friction at the grammar layer and leaves the identity friction untouched.
The Hidden Tax
Time is the real cost. Rewrite cycles. Re-reads. The mental energy spent worrying about how you sound rather than what you are saying.
A non-native speaker in Slack isn't looking for a grammar checker. They're looking for a way to sound like themselves, not a watered-down version, not a careful translation, not someone who learned English in a classroom, but someone who thinks and speaks and works in English every day.
That voice exists on calls. It should exist in writing too.
Whisper Dictation and Local Processing
Recitey runs Whisper, the open speech recognition model with 96.3% word accuracy on LibriSpeech, directly on your device. No metering. No word count caps. No subscription tier that stops you at 2,000 words a month.
Speak your message into Slack. The draft appears as you spoke it: your cadence, your tone, your actual voice preserved in text. No variable cost per word. No translation layer. No assumption that you need correction.
It's not about capturing every "um" and "uh." Recitey knows how to polish rough speech into clean writing in under 2 seconds. The difference is what stays: your voice.
Who This Changes
For a senior PM in Stockholm or Berlin, this means the 30-minute Slack message becomes a 90-second draft. For a sales engineer in Barcelona logging notes in Salesforce, it means the note sounds like you, not like someone translating down into English.
The performance review gap, the one about conciseness in writing, doesn't disappear because Recitey is magic. It disappears because the medium stops fighting you.
The problem was never that you couldn't write English. It was that writing English in Slack made you sound like someone else, and you could feel it every time you re-read before sending.