The transcript problem
Say a few lines of a scene out loud into a general dictation tool and you get something like this back:
| What you said | What a plain transcript gives you |
|---|---|
| Maya lowered the pistol comma her hand steady period new paragraph you said the vault was empty comma she said period new paragraph he didn’t answer period | maya lowered the pistol comma her hand steady period new paragraph you said the vault was empty comma she said period new paragraph he didn’t answer period |
The words are correct. Everything else is wrong: the punctuation you spoke is sitting there as the literal words “comma” and “period,” there is one run-on block instead of three lines, the dialogue has no quotation marks, and the new speaker did not get a new paragraph. Before it is fiction, you have to reshape all of that by hand — and the reshaping eats the time the dictation just saved.
What “a scene” actually requires
Turning that block into prose is not one step. It is several, and they compound:
- Spoken punctuation becomes real punctuation. “Comma” becomes a comma; “new paragraph” becomes a line break, not the words.
- Sentences get capitals and spacing. The first word of each sentence is capitalised; stray filler like “um” is dropped.
- Dialogue gets quoted. What a character says is wrapped in quotation marks, with punctuation tucked inside them the way fiction expects.
- A new speaker gets a new paragraph. This single rule is what makes a page of dialogue readable, and it is exactly what plain transcripts miss.
- Your names are spelled your way. “Maya,” not “Maia”; your invented city, not the closest real-world guess.
Done by hand, that is a full editing pass on every take. Done well, it is invisible — the text simply arrives correct.
How Murmurtale does it as you talk
Murmurtale runs the reshaping automatically, on your Mac, in the moment. After it transcribes your speech, a cleanup stage applies the punctuation, capitalisation, and spacing rules, and an on-device pass labels each sentence as narration, spoken dialogue, or interior thought — then quotes the dialogue, breaks a paragraph when the speaker changes, and sets interior thought apart. Names come from dictionaries you control, so your world stays spelled correctly. The finished text lands at your cursor in the app you are already writing in.
The same lines above come through like this:
Maya lowered the pistol, her hand steady.
“You said the vault was empty,” she said.
He didn’t answer.
This is an illustrative example of Fiction-mode cleanup, not a screenshot. The exact result depends on what you dictate and your settings.
Cleanup, not invention
There is a line Murmurtale does not cross: it reshapes what you said, and it never writes new sentences. It will fix your punctuation and format your dialogue; it will not continue your scene, invent description, or decide what a character says next. Even the strongest cleanup level smooths sentences you dictated rather than adding ones you didn’t. You are the author; the tool is just doing the retyping you would otherwise do yourself.
For the mechanics of how the dialogue pass decides what to quote, see Dialogue and paragraphs in the Help Center.
Where to go next
- Your first dictation — set up and speak your first line.
- Modes and hotkeys — Fiction vs Plain, and the keys that pick them.
- Local-first dictation — where your voice and text actually go.