Voice Dictation

Record a voice memo from inside a note, save the raw audio clip, and get a diarized transcript inserted directly into the current page — enriched with an LLM summary, [[wiki-links]] to related notes, and tags from your vault.

How it works

mic → Rust capture (FLAC saved in <vault>/assets/recordings/)
        ↓
OpenAI-compatible STT endpoint (configurable, e.g. LocalAI)
        ↓  segments + speaker labels
Stratum pipeline
  ├─ summary  → your configured LLM provider (same one as AI features)
  ├─ links    → related-note candidates → LLM picks [[wiki-links]]
  ├─ tags     → existing vault tags → LLM picks 1–3
  ↓
Memo markdown inserted at the end of the current note
  ↓
reindexed → backlinks, search, graph and tag cloud pick it up

• The whole clip is kept as FLAC under assets/recordings/ and linked from the memo (🔊 Listen to recording).
• Diarization separates speakers when the endpoint supports it (POST /v1/audio/diarization). If the endpoint has no diarization endpoint, dictation degrades to a plain transcript.
• Summary generation uses the same LLM provider configured in Settings → AI, so nothing extra to set up.
• Backlinks & tags are never invented: the LLM may only choose from existing notes/tags found in your vault, so every [[link]] resolves.

Recording

1. Open a note and press the microphone icon in the page header.
2. Press Record voice memo. A recording chip with a live timer appears.
3. Press Stop & save — the clip is written to assets/recordings/YYYY-MM-DD_HHMMSS_<page>.flac.
4. Choose options and press Transcribe:
5. Diarize speakers — split the transcript into per-speaker turns.
6. Summarize — prepend an LLM summary block.
7. Identify voices — match speakers against enrolled voices (see below).
8. The memo is appended to the note: header with clip link, summary, **Speaker 1:** … turns, Related: links and #tags.

You can also Cancel a recording to discard the clip.

Assigning names to voices

After transcription, the panel lists every detected speaker. Give each one a name:

• Assign — the memo is re-rendered with the name (e.g. **Alice:** …). The name is remembered in the vault.
• Assign + enroll voice — the same, plus a short sample of that speaker's voice is embedded and stored. Future recordings will auto-identify this person's turns by voice.

Voices are stored in <vault>/.pkm/speakers.toml (names + embeddings) with the reference clips under assets/speakers/.

Configuration

Settings → AI → Voice Dictation (STT):

• Transcription endpoint — any OpenAI-compatible /v1/audio/transcriptions server (e.g. a local LocalAI instance). Leave empty to disable dictation.
• Transcription model — e.g. whisper-1.
• Diarization model — used for speaker separation when available (e.g. pyannote-diarization, a sherpa-onnx pyannote-3.0 setup).
• Language hint — optional (en, de, …); empty = auto-detect.
• Defaults for the per-memo toggles (diarize / summarize / identify).
• Test Connection — verifies the endpoint answers and lists models.

Equivalent config file entries (.pkm/config.toml):

[stt]
endpoint = "http://localhost:8081"
# api_key = "..."          # optional
model = "whisper-1"
diarize_model = "pyannote-diarization"
language = "en"            # optional
diarize = true
auto_summarize = true
auto_identify = true
use_llm_gateway_and_auth = false   # optional; reuse the AI endpoint+key for STT

Reuse the LLM gateway

If your transcription server is the same OpenAI-compatible gateway you configured in Settings → AI, tick Use gateway and auth from LLM in Settings → AI → Voice Dictation instead of setting a separate transcription endpoint and key. Transcripts, diarization, speaker recognition and Test Connection then use the AI endpoint and API key. See AI Features → Reusing the LLM gateway and auth.

::: tip Keep the STT key out of the vault Set the STRATUM_STT_API_KEY environment variable instead of writing an api_key into .pkm/config.toml. A configured env var takes precedence; an empty or whitespace-only env var is ignored and the config value (if any) is used instead. :::

Setting up a diarization model (LocalAI example)

Any endpoint that answers POST /v1/audio/diarization works. On LocalAI the smallest reliable option is pyannote-3.0 via the sherpa-onnx backend (two ONNX models, ~35 MB total — no GPU needed):

# /models/pyannote-diarization.yaml on the LocalAI server
name: pyannote-diarization
backend: sherpa-onnx
type: diarization
parameters:
  model: sherpa-onnx-pyannote-segmentation-3-0/model.onnx
options:
  - diarize.embedding_model=3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx
  - diarize.threshold=0.5
  - diarize.min_duration_on=0.3
  - diarize.min_duration_off=0.5
known_usecases:
  - FLAG_DIARIZATION

Download the two models from the sherpa-onnx releases (speaker-segmentation-models → sherpa-onnx-pyannote-segmentation-3-0, speaker-recongition-models → 3dspeaker_speech_campplus_sv_zh-cn_16k-common.onnx) into the server's models directory, install the sherpa-onnx backend (/local-ai backends install sherpa-onnx) and restart.

The summary/links/tags steps reuse the main AI provider; if it is not configured those steps are skipped and you get the plain transcript.

One-Tap Capture into Today's Journal

In addition to the full dictation panel (used on an individual note), the Journal offers a one-tap voice-capture button that records straight into today's journal:

• Mobile — a fixed microphone FAB (bottom-right). Tap to record, tap the stop button to save & transcribe, or use the cancel button to discard.
• Desktop — a "Record voice memo" button in the journal header.

The flow reuses the exact same transcription pipeline and memo rendering as the dictation panel — nothing is reimplemented. After a successful transcription the journal editor reloads automatically so the inserted memo is visible immediately. The button then returns to the idle "record" state after a short success indicator, so you can record another memo without leaving the journal (one-tap flow is repeatable).

Permissions

• Desktop — the microphone is trivially available; no prompt is shown.
• Android — a system microphone-permission dialog is requested the first time you record. If you deny it, a banner appears with an Enable action.
• iOS — the system permission prompt appears on the first audio stream (NSMicrophoneUsageDescription is declared by the build).

Error fallback (data never lost)

If the transcription endpoint is unreachable or fails, the saved clip is not discarded: a placeholder memo referencing the recording is inserted into the journal and a Retry transcription action is shown. If no transcription endpoint is configured at all, the recording is still saved and a Configure STT prompt links you to Settings → AI.

Enabling / disabling

The one-tap button can be toggled under Settings → AI → Voice-Capture Button. When disabled, the button is hidden — the full dictation panel is still available on any individual note.