Skip to content

Voice-to-text dictation app with local (Nvidia Parakeet/Whisper) and cloud models (BYOK). Privacy-first and available cross-platform.

fix(notes): cap live speaker count to expected count in note recordings (#967)

Gabriel Stein533f9e47306c
diff --git a/src/helpers/diarization.js b/src/helpers/diarization.js
index 5d962f248..bbca06802 100644
--- a/src/helpers/diarization.js
+++ b/src/helpers/diarization.js
@@ -8,7 +8,7 @@ const { resolveBinaryPath, gracefulStopProcess } = require("../utils/serverUtils
const { getModelsDirForService } = require("./modelDirUtils");
const { convertToWav } = require("./ffmpegUtils");
const { getSafeTempDir } = require("./safeTempDir");
-const { applyProvisionalSpeaker, applyConfirmedSpeaker } = require("./speakerAssignmentPolicy");
+const { applyConfirmedSpeaker } = require("./speakerAssignmentPolicy");
const sidecarPidFile = require("./sidecarPidFile");
const {
transcriptsOverlap,
@@ -459,8 +459,6 @@ class DiarizationManager {
return null;
};
- let fallbackSpeakerIndex = speakerSet.size;
-
return deduped.map((seg, index) => {
const enriched = { ...seg };
@@ -505,12 +503,6 @@ class DiarizationManager {
speaker: speakerMap.get(bestSpeaker) || bestSpeaker,
speakerIsPlaceholder: false,
});
- } else if (!enriched.speaker) {
- applyProvisionalSpeaker(enriched, {
- speaker: `speaker_${fallbackSpeakerIndex}`,
- speakerIsPlaceholder: true,
- });
- fallbackSpeakerIndex += 1;
}
}
diff --git a/src/helpers/ipcHandlers.js b/src/helpers/ipcHandlers.js
index d21ad1936..97cb23397 100644
--- a/src/helpers/ipcHandlers.js
+++ b/src/helpers/ipcHandlers.js
@@ -7648,7 +7648,9 @@ class IPCHandlers {
);
this.activeMeetingSpeakerConfig = { enabled, expectedCount };
liveSpeakerIdentifier.setEnabled(enabled);
- liveSpeakerIdentifier.setMaxSpeakers(expectedCount);
+ // Live identification only labels other speakers (the mic track is "you"),
+ // so cap at expectedCount - 1 to match resolveSessionMaxSpeakers().
+ liveSpeakerIdentifier.setMaxSpeakers(Math.max(1, expectedCount - 1));
return { success: true };
} catch (error) {
return { success: false, error: error.message };
diff --git a/src/helpers/windowManager.js b/src/helpers/windowManager.js
index 9f4e69373..7d9d2dfde 100644
--- a/src/helpers/windowManager.js
+++ b/src/helpers/windowManager.js
@@ -1021,11 +1021,7 @@ class WindowManager {
showDictationPanel(options = {}) {
const { focus = false } = options;
if (this.mainWindow && !this.mainWindow.isDestroyed()) {
- const wasHidden = !this.mainWindow.isVisible() || this.mainWindow.isMinimized();
-
- if (wasHidden) {
- this._repositionToCursorDisplay();
- }
+ this._repositionToCursorDisplay();
if (this.mainWindow.isMinimized()) {
this.mainWindow.restore();
diff --git a/src/stores/meetingRecordingStore.ts b/src/stores/meetingRecordingStore.ts
index 2ae6aec89..38df199ec 100644
--- a/src/stores/meetingRecordingStore.ts
+++ b/src/stores/meetingRecordingStore.ts
@@ -537,6 +537,16 @@ function reserveSpeakerIndex(speakerId?: string) {
nextPlaceholderSpeakerIndex = Math.max(nextPlaceholderSpeakerIndex, idx + 1);
}
+// Other-speaker cap is expectedCount - 1 (the mic track is "you"); mirrors the
+// backend cap so live labels can't climb past the count the user expects.
+function mintPlaceholderSpeakerId(): string {
+ const expected = useMeetingRecordingStore.getState().sessionExpectedCount;
+ const cap = Math.max(1, expected - 1);
+ const index = Math.min(nextPlaceholderSpeakerIndex, cap - 1);
+ nextPlaceholderSpeakerIndex = Math.max(nextPlaceholderSpeakerIndex, index + 1);
+ return `speaker_${index}`;
+}
+
function assignProvisionalSpeaker(segment: TranscriptSegment): TranscriptSegment {
if (segment.source !== "system" || segment.speaker) return segment;
@@ -584,8 +594,7 @@ function assignProvisionalSpeaker(segment: TranscriptSegment): TranscriptSegment
});
}
- const speakerId = `speaker_${nextPlaceholderSpeakerIndex}`;
- nextPlaceholderSpeakerIndex += 1;
+ const speakerId = mintPlaceholderSpeakerId();
return normalizeTranscriptSegment({
...segment,
@@ -888,9 +897,13 @@ export async function startRecording(args: StartRecordingArgs): Promise<void> {
} else {
useMeetingRecordingStore.setState({ systemPartial: data.text });
if (!systemPartialSpeakerIdValue) {
- const speakerId = `speaker_${nextPlaceholderSpeakerIndex}`;
- nextPlaceholderSpeakerIndex += 1;
- setSystemPartialSpeakerIdentity(speakerId, null);
+ // Reuse the recent system speaker before minting — the partial id is
+ // cleared after every final, so always minting spawned one per utterance.
+ const carried = getRecentSystemSpeaker(Date.now());
+ setSystemPartialSpeakerIdentity(
+ carried?.speakerId ?? mintPlaceholderSpeakerId(),
+ carried?.speakerName ?? null
+ );
}
}
return;