1. Start in the Journal when the destination is not clear yet
Logseq’s official guide says the app opens on the Journals page and recommends it as the default place to capture information, especially while learning the outliner. A new dated Journal is created each day, so it is a useful inbox for a spoken thought whose permanent page or project is not yet obvious.
Open today’s Journal, create or select a block, and confirm the insertion point before starting speech input. Dictation writes into the focused block; it does not decide where the idea belongs, create backlinks, or move it into a project.
- Use today’s Journal for quick capture.
- Confirm the active graph and date before a long dictation.
- Link, move, or refactor the block only after reviewing the words.
2. Dictate prose into one focused Logseq block
Place the insertion point inside an editable block and trigger the shortcut currently shown under System Settings → Keyboard → Dictation. Apple describes Dictation as a system text-input feature for an app’s focused editable field; the current Mac settings show the language, microphone, and processing disclosure that apply.
Speak a short section, stop, and verify names, dates, numbers, punctuation, and the intended block. Logseq also recognizes navigation key combinations when its editor is not focused, so visible insertion focus matters before you start speaking.
3. Keep block boundaries and graph syntax explicit
Logseq is an outliner, not a plain continuous document. Its official shortcuts distinguish a new block from a new line, and its Markdown format includes Logseq-specific structures such as `[[page references]]`, `((block references))`, properties, embeds, task markers, and priorities.
Dictate the natural-language content first. Then use Enter, Shift-Enter, the keyboard, autocomplete, or the appropriate Logseq command to create the exact block hierarchy and syntax. A spoken phrase that sounds like a command is still ordinary text unless the active speech tool and Logseq editor actually interpret it that way.
4. Add tasks, priorities, and dates with Logseq controls
Logseq documents built-in task workflows such as `LATER → NOW → DONE` and `TODO → DOING → DONE`. It also supports priorities and Deadline or Scheduled commands. These markers affect queries and task state, so they deserve deliberate input rather than a best-effort transcript.
After dictating the task description, use Command-Enter to cycle the configured workflow marker or type the relevant slash command. Add priorities, scheduled dates, and deadlines through Logseq’s controls, then confirm the resulting block before relying on a query or reminder.
5. Do not transfer file-graph claims to the DB version
Logseq’s file-graph documentation describes graphs built from local Markdown files, including `journals`, `pages`, and `assets` folders. The project’s current README separately describes the newer DB version as beta and warns that data loss is possible, recommending a dedicated test graph and backups.
The focused-block dictation workflow can be tested in either interface, but storage, backup, synchronization, and migration claims depend on the graph type and build you actually use. Do not call every Logseq graph a folder of Markdown files, and do not use speech input as a reason to skip the DB version’s backup warning.
6. Separate local graph storage from speech processing and sync
A file graph can keep Markdown or Org content in a local folder, but that fact does not identify how microphone audio was processed. Apple tells users to inspect the Dictation disclosure for the current Mac, language, and configuration. A third-party recognizer has its own model, network, retention, and account boundary.
Synchronization is another layer. Logseq documents its own sync option and notes that file graphs can also be synchronized with file-sync services. Once the resulting text is saved or synchronized, it follows the graph’s storage, backup, sharing, and conflict rules regardless of where speech recognition ran.
7. Treat a voice plugin as a separate desktop integration
Logseq describes plugins as third-party tools available on desktop, with their own settings and update lifecycle. A voice or transcription plugin may offer useful capture, but it is not evidence that Logseq itself ships a universal Mac dictation feature.
Before enabling one, inspect its maintainer, source, release history, requested API key, microphone access, speech provider, network use, model downloads, data retention, and graph permissions. Test it in a non-critical graph before trusting it with confidential notes or the current DB beta.
8. Troubleshoot the speech path before rebuilding the graph
If no text appears, test the same Dictation shortcut in TextEdit. If it fails there too, inspect Keyboard → Dictation, the selected language, shortcut, microphone source, Voice Control interaction, and current processing requirement.
If TextEdit works, return to Logseq, open a plain Journal or page block, click until the insertion point appears, and try one short sentence. Temporarily disable conflicting plugins for the test. A sync conflict, graph-indexing issue, command shortcut, or wrong active block can be real without being a speech-recognition failure.
9. Add IraVoice only for local cross-app text input
IraVoice 0.7.1 is a separate option for macOS 26+ Apple-silicon Macs when speech recognition should run on the Mac or one held-key workflow should supply text across supported Logseq, browser, email, note, and developer fields. Models download on first use; the full trial needs no activation, and paid continuation needs one Gumroad key check.
IraVoice is not a Logseq plugin. It does not open or read graphs, choose pages, create block hierarchies, write page references, set task state, run slash commands, manage plugins, configure sync, migrate graph formats, or make backups. It targets a supported focused editable field and falls back to the clipboard when direct insertion is uncertain.