Drafts at thinking pace.

Why writers reach for dictation

Three numbers from cognitive science: humans think at roughly 250 WPM (Brysbaert/Rayner 2019), speak at 150 WPM in conversation, and type at 40 WPM on average. Typing is so much slower than thinking that you're constantly waiting for your hands to catch up — which causes you to forget the back half of the sentence you'd already worked out, which makes drafts feel laboured.

Dictation closes most of that gap. You're not as fast as pure thought, but you're a lot closer than you were typing. The catch is that most dictation apps render verbatim and you have to manually clean up filler words, false starts, and run-on sentences afterwards. Neuron Flame's AI Enhancement layer handles that cleanup automatically — with a prompt you control.

Three workflows that actually work

1. The "blast draft, polish later" pattern

Hold the hotkey. Talk through the whole section in one go — false starts, repetitions, "actually no, scratch that, what I meant was…" included. Release. Neuron Flame's default prompt cleans the filler words, removes false starts, fixes obvious grammar. You get a clean first draft in 10% of the time it would have taken to type.

2. The "ESL writer's edge"

If you think and speak in English fluently but write English with effort, dictation helps enormously. Set the prompt to "preserve voice and meaning, fix grammar to native-speaker quality, don't change vocabulary unless something is genuinely wrong". The result reads like you, polished.

3. The "interview to article" loop

Drop the recording into the Transcribe Audio tab → get the raw transcript → use the History tab to reach the text → dictate your commentary directly between paragraphs while the source is fresh in your head.

Per-app personalities

Power Mode lets you load different prompts in different apps. Some examples:

• Email mode — formal greeting, professional tone, sign-off. Triggered by Mail.app, Outlook, Spark.
• Notes mode — bullet-formatted thinking, preserve voice. Triggered by Notes, Obsidian, Bear.
• Article mode — fully-formed prose, vary sentence length, eliminate filler. Triggered by Ulysses, Scrivener, Google Docs in Safari.
• DM mode — short, casual, one-line. Triggered by Slack, Messages.

What you keep

Your raw transcript is always saved alongside the AI-cleaned version. If you don't like what the model did, you have the original. If the model fails (rate limit, key invalid, network), Neuron Flame falls back to the raw transcript automatically. You never lose what you said.