fix(transcribe): select multilingual model before download (#2303)

packages/cli/src/whisper/transcribe.test.ts

@@ -1,6 +1,7 @@
 import { describe, expect, it, test } from "vitest";
 import {
   dtwPresetForModel,
+  initialModelForLanguage,
   isWhisperTimeoutError,
   resolveAudioPreparationTimeoutMs,
   resolveWhisperTimeoutMs,
@@ -245,3 +246,21 @@ describe("resolveAudioPreparationTimeoutMs", () => {
     expect(resolveAudioPreparationTimeoutMs(Number.NaN)).toBe(120_000);
   });
 });
+
+describe("initialModelForLanguage", () => {
+  test("keeps the default English-only model when no language is specified", () => {
+    expect(initialModelForLanguage("small.en", undefined)).toBe("small.en");
+  });
+
+  test("uses the multilingual model before downloading for explicit German", () => {
+    expect(initialModelForLanguage("small.en", "de")).toBe("small");
+  });
+
+  test("keeps English-only models for English locale variants", () => {
+    expect(initialModelForLanguage("small.en", "en-US")).toBe("small.en");
+  });
+
+  test("keeps an already multilingual model unchanged", () => {
+    expect(initialModelForLanguage("large-v3", "de")).toBe("large-v3");
+  });
+});

packages/cli/src/whisper/transcribe.ts

@@ -386,6 +386,14 @@ export function dtwPresetForModel(model: string): string {
   return model.replace(/-/g, ".");
 }
 
+export function initialModelForLanguage(model: string, language?: string): string {
+  const baseLanguage = language?.trim().toLowerCase().split(/[-_]/, 1)[0];
+  if (baseLanguage && baseLanguage !== "en" && model.endsWith(".en")) {
+    return model.slice(0, -3);
+  }
+  return model;
+}
+
 /**
  * Transcribe an audio or video file and save transcript.json to the output directory.
  */
@@ -395,7 +403,7 @@ export async function transcribe(
   outputDir: string,
   options?: TranscribeOptions,
 ): Promise<TranscribeResult> {
-  const model = options?.model ?? DEFAULT_MODEL;
+  const model = initialModelForLanguage(options?.model ?? DEFAULT_MODEL, options?.language);
 
   // 1. Ensure whisper binary
   options?.onProgress?.("Checking whisper...");