Voice dictation without cleanup is just transcription — you'd still have to type-edit every "um" and rephrase. Pithflow's AI cleanup pass runs after every dictation and lands you a clean draft on the first try.
Six things, in order: (1) Removes filler words like "um," "uh," "like," "you know," "basically." (2) Applies your most recent self-correction ("actually, scratch that, let me start over" cuts the previous attempt). (3) Adds punctuation and capitalization. (4) Formats lists when you say "first... second... third...". (5) Matches the selected tone (Casual, Professional, Slack, Email, Code comment, Note, Concise, Friendly). (6) Preserves your input language — never translates Spanish to English.
It doesn't paraphrase your words into something generic. It doesn't add information you didn't say. It doesn't "improve" your wording in ways that lose your voice. The cleanup is mechanical — strip the filler, fix the punctuation, apply the tone — not creative.
Cleanup is on by default. Toggle it off in the Style tab if you want raw verbatim output — useful for captioning recorded video, transcribing interviews verbatim, or coaching your own presentation style. The raw transcript is always preserved in your local history, so even with cleanup on you can recover the verbatim version later.
The same raw transcript will be cleaned up differently per tone. "Um so we should probably ship Friday" becomes: • Casual: "We should probably ship Friday" • Professional: "I recommend we ship on Friday." • Slack: "Should we ship Friday?" • Email: "I'd like to propose shipping on Friday." • Code comment: "Ship Friday"
If you're using voice because typing causes pain (RSI, carpal tunnel), every minute spent editing voice output adds back the typing you were trying to avoid. AI cleanup closes ~80% of the gap between "raw transcript" and "ready to ship" — the remaining 20% might still need touch-ups, but it's a fraction of the keystrokes.