Skip to content

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

fix: respect sttConfig dictation mode for realtime models

Gabriel Stein51d105f78714
diff --git a/src/helpers/audioManager.js b/src/helpers/audioManager.js
index f64c471ac..1617c9421 100644
--- a/src/helpers/audioManager.js
+++ b/src/helpers/audioManager.js
@@ -2026,6 +2026,12 @@ registerProcessor("pcm-streaming-processor", PCMStreamingProcessor);
const s = getSettings();
if (s.useLocalWhisper) return false;
+ // For dictation/agent: respect sttConfig mode from the API — this allows
+ // batch mode even for realtime-capable models (e.g. gpt-4o-mini-transcribe).
+ if (this.context !== "notes" && this.sttConfig?.dictation?.mode === "batch") {
+ return false;
+ }
+
if (REALTIME_MODELS.has(s.cloudTranscriptionModel)) {
if (s.cloudTranscriptionMode === "byok") return !!s.openaiApiKey;
if (s.cloudTranscriptionMode === "openwhispr") return !!(isSignedInOverride ?? s.isSignedIn);