diff --git a/apps/mana/apps/web/src/lib/voice/transcribe.ts b/apps/mana/apps/web/src/lib/voice/transcribe.ts
new file mode 100644
index 000000000..99290f65b
--- /dev/null
+++ b/apps/mana/apps/web/src/lib/voice/transcribe.ts
@@ -0,0 +1,52 @@
+/**
+ * Shared voice transcription helper.
+ *
+ * All modules that need speech-to-text use this single function instead
+ * of inlining the fetch call. It posts the audio blob to the SvelteKit
+ * proxy at /api/v1/voice/transcribe, which forwards to mana-stt.
+ */
+
+export interface TranscribeResult {
+	text: string;
+	language: string | null;
+	durationSeconds: number | null;
+	/** STT backend/model identifier returned by mana-stt (e.g. "whisperx-large-v3"). */
+	model: string | null;
+}
+
+/**
+ * Transcribe an audio blob via the server-side STT proxy.
+ *
+ * @throws on HTTP errors or network failures — callers are expected to
+ *   handle errors in a module-specific way (update status, show toast, etc.).
+ */
+export async function transcribeAudio(blob: Blob, language?: string): Promise<TranscribeResult> {
+	const form = new FormData();
+	const ext = blob.type.includes('webm') ? '.webm' : blob.type.includes('mp4') ? '.m4a' : '.audio';
+	form.append('file', blob, `voice${ext}`);
+	if (language) form.append('language', language);
+
+	const response = await fetch('/api/v1/voice/transcribe', {
+		method: 'POST',
+		body: form,
+	});
+
+	if (!response.ok) {
+		const text = await response.text();
+		throw new Error(text || `HTTP ${response.status}`);
+	}
+
+	const result = (await response.json()) as {
+		text: string;
+		language?: string;
+		durationSeconds?: number;
+		model?: string;
+	};
+
+	return {
+		text: (result.text ?? '').trim(),
+		language: result.language ?? null,
+		durationSeconds: result.durationSeconds ?? null,
+		model: result.model ?? null,
+	};
+}