The shape of the day
The Bureau of Labor Statistics counted 135,400 employed writers and authors in the US as of May 2024, with the field projected to grow 4% through 2034. But those numbers only capture a slice of the real population - they miss the 3.5 million self-published titles released in 2025 alone, a figure that grew 43.5% since 2022 according to BestWriting.com's 2024 writing statistics. The majority of people who write for a living are increasingly self-directed: 77% identify as self-published or hybrid authors, and among full-timers, 62% spend an additional 6+ hours per week on marketing beyond the writing itself.
A working author's day rarely looks like "sit down, write." The sustainable daily output for professional authors clusters around 500-2,000 words of core creative work - Stephen King targets 2,000 words/day; Graham Greene produced 25 novels on a disciplined 500 words/day. Research by Anders Ericsson on expert performers suggests the brain handles roughly 4 hours of deliberate, high-focus work daily before quality degrades (per Aalto University / ScienceDaily coverage of the Ericsson deliberate-practice research). Which means the actual writing time is finite, and everything that interrupts or blocks that window has a real cost.
And then there's the other problem: the work that happens when you're not at a desk. A peer-reviewed study by Schooler et al. (2019, Psychological Science) found that roughly 20% of significant ideas among professional writers occurred during non-work activities or while thinking about unrelated topics. Brain imaging research shows spontaneous mind-wandering occurs about 47% of waking hours. When the brain's default mode network activates during low-demand tasks - walking, showering, driving - "executive processes diminish and associative processes amp up," as neuroscientist John Kounios put it in National Geographic. That's the neurological reason you get your best chapter ideas in the shower. It's also the reason they disappear before you can get to a keyboard.
Where typing slows you down
There are two distinct bottlenecks in a writer's workflow, and they show up at opposite ends of the day.
The first is capture. The ideas that arrive away from a keyboard are often the good ones - the structural insight about chapter 12 that surfaces during a walk, the opening line that arrives while making coffee, the character detail that appears unbidden during a commute. Writers have invented an impressive range of workarounds for this: texting themselves, calling their own voicemail, carrying bath crayons for shower ideas, keeping a notebook in every jacket. These methods work, imperfectly. The writing community has documented the capture problem extensively precisely because no single tool dominated the space.
The second bottleneck is wrist endurance. Writing is one of the heaviest-use keyboard occupations there is. A study documented a 28.4% RSI incidence rate among keyboard workers, and research on office workers with 3-5 hours of daily computer use shows musculoskeletal disorder prevalence ranging from 40% among college students to over 70% among university staff. Writers sit in the heaviest-use tier of that group. Carpal tunnel syndrome affects an estimated 4.8 million US workers at any given time (per CDC / NIOSH, 2010 National Health Interview Survey), and when a writer develops it, their primary output mechanism is impaired with no easy substitute if they rely entirely on typing.
The math on daily word counts makes this concrete. At 40 words per minute - a realistic sustained typing pace for prose, not sprint (per Aalto University / ACM CHI 2018 keystroke study) - producing 1,500 words of draft requires about 37 minutes of pure typing. That sounds manageable until you add editing passes, email, social posts, research notes, and query letters. The hands accumulate hours, not minutes. The body notices.
How voice fits this workflow
Voice isn't a replacement for every part of writing. There's a specific division of labor where it earns its place, and being honest about it is more useful than overselling it.
Where voice works well for writers:
Idea capture on the go. When you're walking and the right scene transition arrives, speaking it into your phone takes 15 seconds. Fishing for a pen, unlocking a notes app, or trying to hold the sentence in working memory until you reach a desk all carry a real loss rate. Speaking it out loud is the lowest-friction option that actually produces a saved record.
First-draft dictation. This is where the speed math matters most. A Stanford/UW/Baidu study (Ubicomp 2018) found that speech input on a mobile device ran 3x faster than touchscreen typing - roughly 150 words per minute versus 40 WPM for the average typist. That gap is smaller at a full keyboard, but it still exists. More importantly, voice forces a different cognitive mode: you can't backspace and self-edit mid-sentence, which for some writers produces more fluid first drafts. You're generating, not revising.
Outlining and structure work. Talking through "okay, chapter 9 needs to open with the argument, then the flashback, then the return" is something many writers already do out loud to themselves. Capturing that out loud into a tool rather than losing it is a small habit change with compounding returns.
Notes to self between sessions. The observation about a character's motivation that occurs to you while making lunch. The research question you need to track down. The line you want to cut but aren't sure about yet. These are low-stakes captures that don't need to be precious - they need to be saved.
Where voice doesn't fit:
Voice dictation produces a first draft, not a final one. The output after AI cleanup is clean and grammatically tidy, but it's still prose that was spoken, and spoken prose has a different rhythm than written prose. The last polish pass - sentence-level revision, paragraph pacing, word choice - that still happens at the keyboard. Voice is for generating material, not for final craft decisions.
Voice also doesn't help with structured work that requires precise formatting: a table of contents with specific section lengths, a spreadsheet tracking submissions, a query letter that needs exact wording from a publisher's guidelines. Those need typing.
AICHE specifically for writers and authors
These are the AICHE features that actually matter for this workflow, and why.
Cross-device sync. You capture an idea on your phone during a walk, sit down at your Mac an hour later, and the note is there. No AirDrop, no email to yourself, no copy-paste. The sync is end-to-end encrypted - notes are encrypted on your device before syncing, with a key only you hold. For writers who work across a phone and a desktop (which is most of them), this is the feature that makes mobile capture worth doing consistently rather than as an occasional hack.
Apple Watch recording. Tap your wrist and talk, no length cap. The recording transcribes and syncs to your iPhone and every other AICHE device. This is genuinely useful for the subset of ideas that arrive during a run, a walk, or any other activity where your phone is in your pocket. It lowers the friction of capture to the lowest possible point: you don't have to stop what you're doing and take out a device. Note: Watch recording syncs via iPhone, so your iPhone needs to be nearby.
Android home-screen widget. One tap on the widget starts recording. No opening the app, no navigating to a new note screen. For Android users, this is the equivalent capture shortcut - something you can activate mid-thought without stopping to launch anything.
AI cleanup. After you speak, AICHE removes filler words (um, uh, like), false starts, and repetitions, then adds punctuation and paragraph breaks. The cleanup runs through a post-processing pipeline that also handles hallucination filtering and applies your custom vocabulary. The result isn't polished prose, but it's not raw transcript either - it's readable draft material that you can work with. For 15 minutes of audio, this takes roughly 3 seconds.
Custom vocabulary. You can teach AICHE up to 50 custom terms that get enforced on every transcript - character names, invented place names, series titles, publisher names, agent names. Once they're in the vocabulary list, they're spelled correctly every time rather than mangled by autocorrect.
Never-lose offline queue. If the network drops mid-recording, the app crashes, or you are on a plane, the audio saves locally and encrypted on disk. When connectivity returns, the queue processes automatically. For writers who use voice capture specifically in the moments when they're away from a desk (and therefore often on intermittent connectivity), this is the failure mode that kills adoption in other tools. Recording something and losing it, especially a long one, is enough to make most people stop using the tool. AICHE reduces that failure mode by keeping the local queue available until processing can resume.
Desktop global hotkey. On Mac it's Control + Option + R; on Windows and Linux it's Ctrl + Alt + R. Press it in any app, speak, press it again to stop. The text inserts at your cursor in whatever you're writing in - Scrivener, Word, Google Docs, your email client. It's toggle-based, not push-to-hold. For desktop first-draft dictation, this is the main workflow.
Honest tradeoffs
AICHE is not the right tool for every part of a writer's work. Specific cases where it doesn't fit:
Not a creative writing AI. AICHE cleans up what you say; it doesn't generate story content, suggest plot directions, or write paragraphs in your style. If you want AI writing assistance, tools like Claude, ChatGPT, or Sudowrite are built for that. AICHE's job is to get your spoken words into usable text, accurately and without losing anything. It doesn't do the thinking for you.
Voice drafts still need editing. This is true of any dictation workflow, not an AICHE-specific limitation. Spoken prose drafts are rougher than written ones - different sentence rhythms, more run-ons, occasional repetition that survives cleanup. Voice is fast for generating volume; it's not a substitute for revision.
Not for the very last polish pass. Fine-grained word choice, sentence musicality, paragraph-level pacing decisions - those are editing-table work. You wouldn't dictate a final proofreading pass. Voice gets you material to edit; the editing itself stays at the keyboard.
Cloud round-trip. Audio goes from your device to Groq (the named transcription provider), gets processed, and is discarded immediately after processing, within 1 second. AICHE is honest about this - it's not local-only processing. If you have specific concerns about audio leaving your device for any content, that's the honest constraint to weigh. The privacy posture (named provider, fast deletion, zero audio retention, E2EE sync) is strong relative to the category, but it's not the same as a purely local tool.
Not a meeting transcription tool. AICHE doesn't record and transcribe live conversations between multiple speakers. For meeting notes and interview transcription, Otter.ai is built specifically for that.
Desktop app is English-only. The desktop app, Chrome extension, and Obsidian plugin menus are in English only. Voice input itself works in 99 languages on every platform, and the mobile apps (iPhone, iPad, Android) are localized in 28 languages - but if you work on a desktop and your primary language isn't English, the interface is English. This doesn't affect transcription quality, only the UI.
No specialist vocabulary packs. If you write in a domain with dense technical terminology - science fiction with real physics, medical thrillers, legal procedurals - the 50-entry custom vocabulary gets you partway there, but it's not a trained specialist vocabulary like Dragon Professional's medical or legal dictionaries. You'll encounter more transcript corrections in the first few months until you've loaded your key terms.
What to try first
If you want to test whether voice fits your workflow, these three experiments give you useful signal inside a week without changing your whole process.
Experiment 1: Capture one idea per day by voice for five days. Pick a moment each day when an idea arrives away from your desk - the commute, a walk, before bed. Speak it into AICHE on your phone instead of texting yourself or hoping you'll remember it. Don't try to dictate a full scene, just the idea. After five days, look at the notes. Check whether the captures are actually useful versus what you'd normally retain. Most people find the capture-to-discard ratio is better than they expected.
Experiment 2: Dictate one section of a first draft. Pick a section you've been avoiding starting, or one you know well enough to talk through out loud. Press Ctrl+Alt+R (or ⌃+⌥+R on Mac) in your writing app, speak the section, stop. Clean it up in your normal editing pass. This tells you whether voice drafting actually helps your specific writing process - some writers find it faster; others find the output requires so much rewriting that it's slower overall. You need to run the test on your own work to know.
Experiment 3: Use the Watch for one type of idea. If you have an Apple Watch, pick one recurring context where you often have ideas but can't capture them (runs, dishwashing, waiting in line). For one week, tap the Watch and speak every time something usable comes up in that context. You're testing whether the wrist-tap barrier is low enough for that specific context, and whether the ideas that survive a Watch recording are worth having in your notes.
Try AICHE
7-day free trial, no credit card. Personal plan from $3.99/month on annual. Works on macOS, Windows, Linux, iPhone, iPad, Apple Watch, Android, Chrome, and Obsidian.