fix(dub): keep language and media tools ready (#1679)

Fixes #1677 and #1678.

Publishes first-run media tools to the live backend, provides precise cross-platform missing-process guidance, and keeps localized source-language selection available before transcription. Includes regression coverage and deterministic model-store test isolation.

CHANGELOG.md

@@ -56,6 +56,8 @@ the frozen-backend fallback mirror it for their toolchains.
 - The OmniVoice guide now covers combining style attributes with a reference clip (consistent instruct stabilizes cloning; the reference wins conflicts), inline pronunciation control (pinyin / CMU phonemes), and corrects the claim that the default engine can't do voice design — it can, from attributes (#1565)
 
 ### Fixed
+- Dubbing keeps the source-language selector visible after a local file is chosen, so ASR can be pinned before transcription starts (#1678) — thanks @Lonki-lomki-cloud!
+- First-run media-engine downloads become available to TTS immediately without a restart, and missing media-process failures now point to repair controls (#1677) — thanks @farhataligpt-dev!
 - Source installs on AMD GPUs honour `OMNIVOICE_TORCH_VARIANT=rocm`: `bun run desktop` now swaps in the ROCm torch wheel after `uv sync` and launches the backend without re-syncing, instead of silently reverting to the CPU-only CUDA build on every start (#1665) — thanks @uberclokr!
 - `bun run desktop` on a fresh clone no longer fails with "resource path `../../frontend/dist` doesn't exist" — the dev launcher creates the placeholder Tauri resource directory before compiling (#1664) — thanks @uberclokr!
 - macOS no longer loses TTS after the first request when Python lacks `os.waitid`; subprocess ownership now uses a safe `waitpid` fallback without risking reused process groups (#1656) — thanks @paoloantinori!

backend/api/routers/generation.py

@@ -459,6 +459,31 @@ def _is_timeout_failure(e) -> bool:
     return False
 
 
+def _is_media_process_launch_failure(exc: BaseException) -> bool:
+    """Identify an ffmpeg/ffprobe launch ENOENT without guessing from a file name."""
+    if not isinstance(exc, FileNotFoundError):
+        return False
+
+    # A regular missing reference/model file may itself be named "ffmpeg".
+    # Require the innermost raise site to be Python's process launcher so that
+    # basename collisions keep the normal missing-file diagnosis (#1677).
+    traceback_cursor = exc.__traceback__
+    if traceback_cursor is None:
+        return False
+    while traceback_cursor.tb_next is not None:
+        traceback_cursor = traceback_cursor.tb_next
+    origin_module = traceback_cursor.tb_frame.f_globals.get("__name__", "")
+    if origin_module != "subprocess" and not origin_module.startswith("asyncio."):
+        return False
+
+    filename = getattr(exc, "filename", None)
+    if not filename:
+        return "[winerror 2]" in str(exc).lower()
+    return os.path.basename(str(filename)).lower() in {
+        "ffmpeg", "ffmpeg.exe", "ffprobe", "ffprobe.exe",
+    }
+
+
 def _oom_friendly_reraise(e):
     """Best-effort cache flush + the user-facing OOM hint shared by both
     inference paths."""
@@ -483,6 +508,21 @@ def _oom_friendly_reraise(e):
     # that lost its +x bit) is NOT an OOM — don't send the user to the Flush
     # button; tell them what's actually wrong.
     es = str(e)
+    # #1677: Windows CreateProcess reports a missing executable as a bare
+    # ``FileNotFoundError: [WinError 2] ...`` with no filename, while POSIX
+    # includes the missing ffmpeg/ffprobe name. The bundled-media downloader
+    # now republishes PATH as soon as it finishes, but a failed/blocked
+    # download still needs an actionable recovery rather than the unknown-
+    # error dead end. Keep missing reference/model files on their own path.
+    for _exc in _exception_chain(e):
+        if _is_media_process_launch_failure(_exc):
+            raise RuntimeError(
+                "A required media program couldn't be launched. Open "
+                "Settings → Audio tools and use "
+                "Download/Repair for the media engine, then retry. If Audio "
+                "tools is already ready, repair the selected TTS engine and "
+                f"restart VoiceStudio. Underlying error: {_safe_exc_text(_exc)}"
+            ) from e
     if isinstance(e, PermissionError) or "Permission denied" in es or "Errno 13" in es:
         raise RuntimeError(
             f"A required engine binary couldn't be executed (permission denied). "

backend/services/media_tools.py

@@ -154,6 +154,17 @@ def bundled_dir() -> str:
     return os.path.join(media_tools_dir(), f"ffbin-{_FFBIN_COMMIT[:12]}", _platform_key())
 
 
+def _publish_bundled_on_path() -> None:
+    """Make a newly validated bundle visible to bare-name subprocess calls."""
+    directory = os.path.abspath(bundled_dir())
+    current = os.environ.get("PATH", "")
+    entries = current.split(os.pathsep) if current else []
+    if os.path.normcase(directory) in {os.path.normcase(entry) for entry in entries if entry}:
+        return
+    os.environ["PATH"] = os.pathsep.join([directory, *entries])
+    logger.info("Published the acquired media-tool directory on PATH")
+
+
 def _exe(name: str) -> str:
     return f"{name}.exe" if sys.platform == "win32" else name
 
@@ -231,12 +242,21 @@ def acquire_bundled(wait: bool = False) -> dict:
         _ops["acquire"].update(state="running", progress=0.0, error=None)
 
     if all(bundled_tool_path(t) and _binary_runs(bundled_tool_path(t)) for t in TOOLS):
+        # The bundle may have arrived after startup's one-time PATH publish
+        # (first-run acquisition is asynchronous). Make it visible to pydub
+        # and other dependencies that launch ffmpeg/ffprobe by bare name now,
+        # without requiring a backend restart (#1677).
+        _publish_bundled_on_path()
         _set_op("acquire", state="done", progress=1.0)
         return _op_snapshot()["acquire"]
 
     def _worker():
         try:
             _do_acquire()
+            # Startup cannot publish binaries which do not exist yet. The
+            # background worker must complete that second half atomically with
+            # installation so the very next synthesis can use the tools.
+            _publish_bundled_on_path()
             _set_op("acquire", state="done", progress=1.0, error=None)
             logger.info("media-tools: bundled ffmpeg/ffprobe installed at %s", bundled_dir())
         except Exception as e:

frontend/src/components/dub/IdleSkeleton.jsx

@@ -25,7 +25,7 @@ import DubbingDemo from '../DubbingDemo';
 import DubFailureNotice from './DubFailureNotice';
 import PrepOverlay from './PrepOverlay';
 import TranscribeOverlay from './TranscribeOverlay';
-import { LANG_CODES } from '../../utils/languages';
+import { LANG_CODES, languageLabel } from '../../utils/languages';
 
 const SPEAKERS_INPUT =
   'w-[52px] ml-[4px] px-[6px] py-[4px] rounded-[6px] border border-[var(--border,#3c3836)] bg-[var(--input-bg,#282828)] text-inherit text-[12px]';
@@ -59,6 +59,7 @@ function AsrInstallStatus({ t, install, onAbort }) {
 
 export default function IdleSkeleton({
   t,
+  uiLocale,
   dubVideoFile,
   activeProjectName,
   dubFilename,
@@ -272,6 +273,26 @@ export default function IdleSkeleton({
                     />
                   </label>
                 )}
+                <label className="inline-flex items-center gap-[5px] text-[12px] text-[var(--muted,#a89984)] whitespace-nowrap">
+                  <Globe size={13} /> {t('dub.source_language')}
+                  <select
+                    className="input-base text-[0.65rem]"
+                    value={dubSourceLangCode}
+                    disabled={
+                      dubStep === 'uploading' ||
+                      dubStep === 'transcribing' ||
+                      dubStep === 'installing-asr'
+                    }
+                    onChange={(event) => setDubSourceLangCode(event.target.value)}
+                  >
+                    <option value="auto">{t('bootstrap.auto_detect')}</option>
+                    {LANG_CODES.map((language) => (
+                      <option key={language.code} value={language.code}>
+                        {languageLabel(language.code, uiLocale, language.label)} — {language.code}
+                      </option>
+                    ))}
+                  </select>
+                </label>
                 <label
                   className="inline-flex items-center gap-[5px] text-[12px] text-[var(--muted,#a89984)] whitespace-nowrap"
                   title={t('dub.num_speakers_help')}
@@ -485,7 +506,7 @@ export default function IdleSkeleton({
                     <option value="auto">{t('bootstrap.auto_detect')}</option>
                     {LANG_CODES.map((language) => (
                       <option key={language.code} value={language.code}>
-                        {language.label} — {language.code}
+                        {languageLabel(language.code, uiLocale, language.label)} — {language.code}
                       </option>
                     ))}
                   </select>
@@ -506,7 +527,7 @@ export default function IdleSkeleton({
                   >
                     {LANG_CODES.map((lc) => (
                       <option key={lc.code} value={lc.code}>
-                        {lc.label} — {lc.code}
+                        {languageLabel(lc.code, uiLocale, lc.label)} — {lc.code}
                       </option>
                     ))}
                   </select>

frontend/src/pages/DubTab.jsx

@@ -651,6 +651,7 @@ export default function DubTab(props) {
       {showIdleSkeleton && (
         <IdleSkeleton
           t={t}
+          uiLocale={i18n.language}
           dubVideoFile={dubVideoFile}
           activeProjectName={activeProjectName}
           dubFilename={dubFilename}

frontend/src/test/DubIdleSkeleton.test.jsx

@@ -29,6 +29,7 @@ function baseProps(overrides = {}) {
   const noop = vi.fn();
   return {
     t: i18n.t,
+    uiLocale: 'en',
     dubVideoFile: null, // URL-ingest / restored job: no local File
     activeProjectName: '',
     dubFilename: '',
@@ -64,6 +65,8 @@ function baseProps(overrides = {}) {
     youtubeCookieFile: null,
     setYoutubeCookieFile: noop,
     dubLangCode: 'en',
+    dubSourceLangCode: 'auto',
+    setDubSourceLangCode: noop,
     setDubLangCode: noop,
     setDubLang: noop,
     landingAdvOpen: false,
@@ -108,6 +111,29 @@ describe('IdleSkeleton — pipeline-stage vs idle dropzone', () => {
     expect(screen.getByLabelText('Choose a cookies.txt export')).toBeInTheDocument();
   });
 
+  it('keeps source-language selection available after choosing a local file', () => {
+    const setDubSourceLangCode = vi.fn();
+    renderIdle({
+      dubVideoFile: new File(['video'], 'thai.mp4', { type: 'video/mp4' }),
+      setDubSourceLangCode,
+    });
+
+    fireEvent.change(screen.getByRole('combobox', { name: i18n.t('dub.source_language') }), {
+      target: { value: 'th' },
+    });
+    expect(setDubSourceLangCode).toHaveBeenCalledWith('th');
+  });
+
+  it('localizes source-language option names for the active UI locale', () => {
+    renderIdle({
+      dubVideoFile: new File(['video'], 'thai.mp4', { type: 'video/mp4' }),
+      uiLocale: 'fr',
+    });
+
+    const thai = new Intl.DisplayNames(['fr', 'en'], { type: 'language' }).of('th');
+    expect(screen.getByRole('option', { name: `${thai} — th` })).toBeInTheDocument();
+  });
+
   it('clears the native cookie picker when the selection is removed', () => {
     const selected = new File(['# Netscape HTTP Cookie File\n'], 'cookies.txt', {
       type: 'text/plain',

frontend/src/test/modelStoreFilter.test.jsx

@@ -65,6 +65,7 @@ vi.mock('../api/system', () => ({
 }));
 vi.mock('../api/external', () => ({ openExternal: vi.fn() }));
 vi.mock('../components/settings/models/RecoBanner', () => ({ default: () => null }));
+vi.mock('../components/settings/VoicePreviewsPanel', () => ({ default: () => null }));
 // Surface the FILTERED row set (and the empty-state action) without the
 // virtualizer, which yields no rows in jsdom. The filter itself — TanStack
 // global/column filter state driven by the real toolbar — stays real.

frontend/src/test/modelStoreGrouping.test.jsx

@@ -212,6 +212,7 @@ vi.mock('../api/system', () => ({
 }));
 vi.mock('../api/external', () => ({ openExternal: vi.fn() }));
 vi.mock('../components/settings/models/RecoBanner', () => ({ default: () => null }));
+vi.mock('../components/settings/VoicePreviewsPanel', () => ({ default: () => null }));
 // Surface the per-section row set without the virtualizer (yields no rows in
 // jsdom). Section chrome (headers, incompatible toggle) stays real.
 vi.mock('../components/settings/models/ModelsTable', () => ({

frontend/src/test/modelStoreResidency.test.jsx

@@ -150,6 +150,7 @@ vi.mock('../api/setup', () => ({
 }));
 vi.mock('../api/external', () => ({ openExternal: vi.fn() }));
 vi.mock('../components/settings/models/RecoBanner', () => ({ default: () => null }));
+vi.mock('../components/settings/VoicePreviewsPanel', () => ({ default: () => null }));
 // Render every row's cells for real (the virtualizer yields nothing in jsdom).
 vi.mock('../components/settings/models/ModelsTable', () => ({
   default: ({ tableRows }) => (

frontend/src/utils/languages.js

@@ -92,3 +92,13 @@ export const LANG_CODES = [
   { code: 'yo', label: 'Yoruba' },
   { code: 'zu', label: 'Zulu' },
 ];
+
+/** Localized language name, with the static English label as a WebView fallback. */
+export function languageLabel(code, uiLocale = 'en', fallback = code) {
+  try {
+    const names = new Intl.DisplayNames([uiLocale, 'en'], { type: 'language' });
+    return names.of(code) || fallback;
+  } catch {
+    return fallback;
+  }
+}

tests/test_generation_audio_guard.py

@@ -7,6 +7,7 @@ misleading "ran out of memory". Two guards: sanitize non-finite samples before
 any encode, and classify a decode/ffmpeg failure as unreadable-audio (not OOM).
 """
 import os
+import subprocess
 import sys
 
 import pytest
@@ -21,6 +22,16 @@ from api.routers.generation import (  # noqa: E402
 )
 
 
+def _missing_media_process_error() -> FileNotFoundError:
+    """Capture a real process-launch ENOENT so traceback origin is preserved."""
+    env = {**os.environ, "PATH": os.path.join(os.path.dirname(__file__), "missing-bin")}
+    try:
+        subprocess.run(["ffmpeg"], check=False, env=env)
+    except FileNotFoundError as exc:
+        return exc
+    raise AssertionError("missing ffmpeg unexpectedly launched")
+
+
 def test_sanitize_replaces_non_finite_with_silence():
     t = torch.tensor([0.1, float("nan"), float("inf"), -float("inf"), 0.2])
     out = _sanitize_audio(t)
@@ -109,6 +120,37 @@ def test_unknown_error_is_not_labelled_oom():
     assert "Try the Flush button" not in msg
 
 
+def test_bare_windows_missing_process_error_is_actionable():
+    """CreateProcess omits the executable name from WinError 2 on Windows."""
+    err = _missing_media_process_error()
+    err.filename = None
+    err.strerror = "[WinError 2] The system cannot find the file specified"
+    with pytest.raises(RuntimeError) as ei:
+        _oom_friendly_reraise(err)
+    msg = str(ei.value)
+    assert "doesn't recognize" not in msg
+    assert "Audio tools" in msg
+    assert "required media program" in msg
+
+
+def test_posix_missing_media_process_error_is_actionable():
+    """Equivalent missing-tool failures receive the same guidance on POSIX."""
+    err = _missing_media_process_error()
+    with pytest.raises(RuntimeError) as ei:
+        _oom_friendly_reraise(err)
+    msg = str(ei.value)
+    assert "Audio tools" in msg
+    assert "required media program" in msg
+
+
+def test_missing_file_named_like_media_tool_keeps_missing_file_diagnosis():
+    """A reference/model basename collision is not a process-launch failure."""
+    err = FileNotFoundError(2, "No such file or directory", "ffmpeg")
+    with pytest.raises(RuntimeError) as ei:
+        _oom_friendly_reraise(err)
+    assert "Audio tools" not in str(ei.value)
+
+
 @pytest.mark.parametrize("reason", [
     "CUDA out of memory. Tried to allocate 20.00 MiB",
     "MPS backend out of memory (MPS allocated: 8.00 GB)",

tests/test_media_tools.py

@@ -26,6 +26,7 @@ def mt(monkeypatch, tmp_path):
     import core.prefs as prefs
     import services.media_tools as mt_mod
 
+    original_path = os.environ.get("PATH")
     monkeypatch.setattr(prefs, "_PREFS_PATH", str(tmp_path / "prefs.json"))
     monkeypatch.setattr(mt_mod, "media_tools_dir", lambda: str(tmp_path / "media_tools"))
     auth_dir = tmp_path / "path-authorizations"
@@ -47,6 +48,10 @@ def mt(monkeypatch, tmp_path):
     # (test_pitch_stretch_async was the victim). Explicitly drop them.
     for key in _OVERRIDE_KEYS:
         os.environ.pop(key, None)
+    if original_path is None:
+        os.environ.pop("PATH", None)
+    else:
+        os.environ["PATH"] = original_path
 
 
 def _client():
@@ -165,6 +170,32 @@ def test_acquire_installs_when_checksum_and_probe_pass(mt, monkeypatch):
             assert os.access(p, os.X_OK)
 
 
+def test_acquire_publishes_new_binaries_to_the_live_process(mt, monkeypatch):
+    """A first-run background download must become usable without restart."""
+    payload = _make_zip([f"plat/{mt._exe('ffmpeg')}", f"plat/{mt._exe('ffprobe')}"])
+    _patched_bundle(mt, monkeypatch, payload)
+    monkeypatch.setattr(mt, "_binary_runs", lambda p: True)
+    published = []
+    monkeypatch.setattr(mt, "_publish_bundled_on_path", lambda: published.append(True))
+
+    with patch("urllib.request.urlopen", return_value=_FakeResponse(payload)):
+        assert mt.acquire_bundled(wait=True)["state"] == "done"
+
+    assert published == [True]
+
+
+def test_publish_bundled_on_path_prepends_once(mt, monkeypatch):
+    monkeypatch.setenv("PATH", os.pathsep.join(["system-a", "system-b"]))
+
+    mt._publish_bundled_on_path()
+    mt._publish_bundled_on_path()
+
+    entries = os.environ["PATH"].split(os.pathsep)
+    assert entries[0] == os.path.abspath(mt.bundled_dir())
+    assert entries.count(entries[0]) == 1
+    assert entries[1:] == ["system-a", "system-b"]
+
+
 def test_acquire_rejects_binary_that_fails_version_probe(mt, monkeypatch):
     """A checksum-valid download whose binary won't run (wrong arch, corrupt)
     must NOT be installed — the WinError-193 class, caught at install time."""