1. Start with the workflow Scrivener itself documents
Literature & Latte says Scrivener on Mac can use the operating system’s built-in speech recognition. Enable Dictation under System Settings → Keyboard, check the shortcut displayed for the active Mac, then invoke it while an editable Scrivener document has focus.
This is direct text input, not a recording imported into Scrivener and not a Scrivener-owned cloud transcript. A separate recorder or transcription service can produce text that you later paste into the project, but that is a different workflow with its own audio, storage, and processing boundary.
2. Select the exact Binder document before speaking
Open the project and select the chapter, scene, or other document whose text should change. Click in the main editor and verify the insertion point. Dictation supplies words to the focused editable surface; it does not infer the intended Binder item from what you say.
Keep titles, synopses, notes, comments, and manuscript prose distinct. If the cursor is in a metadata field or another editor pane, the same speech can land in the wrong part of the project. Dictate a short test phrase before a long session whenever the focus is uncertain.
- Select the intended Binder item.
- Click the exact editor position.
- Dictate one short test phrase.
- Continue only after the text lands in the correct document.
3. Speak punctuation and structure, then revise
Literature & Latte’s dictation guide treats speech input as a way to get a first draft onto the page. It recommends speaking punctuation and paragraph commands supported by the operating system, then refining the result afterward rather than trying to produce a finished manuscript in one pass.
Review character names, invented places, dialogue punctuation, homophones, scene breaks, numbers, and capitalization. A recognizer can produce grammatical-looking text that changes the story, so visible review remains part of the writing process.
4. Take a snapshot before a major spoken rewrite
Scrivener snapshots preserve a version of an individual document at the moment you create the snapshot. Before replacing a scene, restructuring a chapter, or dictating over selected text, take a snapshot so the earlier document can be compared or restored.
A snapshot is not a full-project backup. Literature & Latte distinguishes document snapshots from Scrivener’s automatic project backups and recommends maintaining proper backups separately.
5. Use Composition Mode for focus, not as a new speech engine
Scrivener’s Composition Mode hides most of the Binder, Inspector, and interface so the current document occupies a distraction-reduced writing surface. That can make a long dictation session easier to follow visually.
Composition Mode changes the writing environment, not the speech processor or destination rules. Confirm the active document before entering the mode, and exit it when you need to verify Binder structure, research, metadata, or neighboring scenes.
6. Keep Dictation separate from Compile
Dictation edits the currently focused project document. Compile is a later Scrivener operation that assembles selected items from the Draft or Manuscript structure and exports a separate file such as PDF or Microsoft Word.
Do not treat successful speech input as proof that the compiled manuscript contains the right scenes, order, section layouts, front matter, or formatting. Review the Binder selection and Compile settings independently, then inspect the exported file before sharing it.
7. Check processing and storage at every layer
For Apple Dictation, read the processing notice under System Settings → Keyboard → Dictation for the active Mac, language, and configuration. Do not infer an offline or on-device guarantee solely because the text appears inside a desktop app.
The resulting prose becomes part of the Scrivener project and its backups. A snapshot adds another document version inside the project; Compile creates another file at the chosen destination. Those saved-text layers are separate from how the microphone audio was processed.
8. Troubleshoot the input path before changing the project
If no words appear, test one sentence in TextEdit with the same Dictation shortcut. Confirm Dictation is enabled, the intended microphone and language are selected, and Voice Control is not taking over the shortcut.
If TextEdit works, return to Scrivener, select an ordinary manuscript document, click in the main editor, and try a short phrase. Text appearing in one editor but not another points to focus, selection, or surface behavior rather than a failed speech recognizer.
9. Add IraVoice only for a distinct local Mac path
Use Apple Dictation first for occasional free speech input into Scrivener. Use Scrivener snapshots, backups, Composition Mode, and Compile for the project-management jobs they already own.
IraVoice 0.7.1 is a separate option for macOS 26+ Apple-silicon Macs when recognition and supported formatting should run on the Mac or a longer spoken outline should become reviewable Markdown first. IraVoice does not open a Scrivener project, choose Binder documents, inspect research, create snapshots, change metadata, reorder scenes, configure Compile, export a manuscript, or save the project.