fix(gallery): OmniVoice Gallery rename, dark dropdowns, and noisy/stale archetype previews (#241)

* fix(gallery): rename to "OmniVoice Gallery" + fix dark-theme dropdown colors

The gallery heading now reads "OmniVoice Gallery" (gallery.title, all 21
locales — brand prefix on each localized word).

The facet filter <select>s (Gender/Age/Pitch/Accent/Language) rendered with
the OS-default light control surface on the dark theme: .facet-select set
background/border from --bg-tertiary / --border-color, which are defined
nowhere. An undefined var() reads as transparent on the sibling <div> filters
(fine over the dark page) but falls back to the native light background on a
form control. Switch to the defined dark-chrome tokens and add
color-scheme: dark + an explicit dark option list so the popup matches across
WebKit / WebView2 / WebKitGTK.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(gallery): archetype previews render a noise buzz instead of voice

The Hype Host, The Podcaster and The Vlogger previews played a loud tonal
buzz, not speech. The preview renderer pinned num_step=16 and seed=42; the
"social" sample script at that exact point lands on a degenerate diffusion
trajectory and collapses to a near-pure tone. The blank-audio guard missed it
because the buzz is loud (peaks near -2 dBFS), not silent — so the garbage was
cached and served. The cache key is (instruct, language) only, so it never
self-corrected.

- Bump preview num_step 16 -> 32: reliably converges to speech across the
  gallery's instruct/script space (one-time, cached render cost).
- Add a spectral-flatness floor (_is_unusable_audio) so a degenerate tonal
  render is rejected like a blank one, reusing the existing retry-on-new-seed
  path. Whisper/breathy voices are broadband (high flatness) so they're safe.

Verified: flatness Hype Host 0.001->0.050, Podcaster 0.0002->0.083,
Vlogger 0.004->0.039; whisper control (Calm Guide) 0.239, not flagged.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* fix(gallery): stop preview playback replaying stale cached audio

Preview audio is re-rendered server-side when an archetype is fixed, but the
URL is stable and the response carried no Cache-Control — so the WebView's
HTTP cache replayed the first clip it ever fetched (e.g. the old buzz)
indefinitely, even after the server file was corrected.

- Frontend: fetch previews with { cache: 'no-store' } so playback always
  pulls current bytes.
- Backend: send Cache-Control: no-cache on the preview response so any client
  revalidates against the ETag instead of serving a stale clip.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

* test(e2e): add Playwright UI smoke + gallery specs and preview-quality unit test

UI testing system to catch regressions like "Use design → Importing a module
script failed" (a dead Vite/module server) and the noisy-preview bug.

- Playwright (frontend/e2e): drives the system chromium (no browser download)
  against the Vite dev server. ui-smoke mounts all 12 routable views and fails
  on any code-split/import failure, uncaught exception, or ErrorBoundary
  fallback. gallery.spec asserts the "OmniVoice Gallery" heading, the dark
  facet dropdowns (computed bg = rgba(255,255,255,0.04), not the OS-default
  light surface), and that opening an archetype in the Designer mounts the
  lazy CloneDesignTab. `bun run e2e`.
- backend/tests/test_archetype_preview_quality.py: unit-tests the
  _spectral_flatness / _is_unusable_audio guard with synthetic signals
  (tone < threshold < speech < noise; loud tone + silence are unusable) and
  pins the render constants. CI-safe — no model/GPU.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>

.gitignore

@@ -129,3 +129,5 @@ marketing.md
 .claude/skills/speckit-*/
 .antigravitycli/
 
+playwright-report/
+.last-run.json

backend/api/routers/archetypes.py

@@ -41,6 +41,17 @@ _PREVIEW_DIR = Path(OUTPUTS_DIR) / "archetype_previews"
 # Seed fixed so repeated renders of the same archetype are reproducible
 # (mirrors scripts/render_demos_omnivoice.py).
 _PREVIEW_SEED = 42
+# Diffusion steps for previews. 16 under-converges: certain (script, seed)
+# points — notably the "social" sample script at seed 42 — collapse to a
+# degenerate tonal buzz (The Hype Host / Podcaster / Vlogger, issue follow-up).
+# 32 reliably converges to speech across the gallery's instruct/script space
+# at a one-time (cached) render cost.
+_PREVIEW_NUM_STEP = 32
+# Spectral-flatness floor below which a render is a degenerate tonal artifact
+# rather than speech. Real, mastered speech sits ~0.04–0.07; a tonal buzz
+# collapses to <0.005. 0.015 separates the two with wide margin and sits well
+# below even breathy/whisper voices (which are broadband → high flatness).
+_DEGENERATE_FLATNESS = 0.015
 
 
 def _preview_key(a: dict) -> str:
@@ -80,6 +91,41 @@ def _is_blank_audio(audio_tensor) -> bool:
         return False
 
 
+def _spectral_flatness(audio_tensor) -> Optional[float]:
+    """Geometric-mean / arithmetic-mean of the power spectrum.
+
+    ~1.0 for broadband noise, →0 for a pure tone. The degenerate diffusion
+    renders this guards against are near-pure tonal buzzes (flatness <0.005),
+    distinct from both silence (caught by ``_is_blank_audio``) and real speech
+    (~0.04+). Returns ``None`` if it can't be computed so callers don't act on
+    a bad measurement.
+    """
+    try:
+        import torch
+
+        t = audio_tensor if isinstance(audio_tensor, torch.Tensor) else torch.as_tensor(audio_tensor)
+        t = t.detach().to("cpu", dtype=torch.float32).flatten()
+        if t.numel() < 1024 or not torch.isfinite(t).all():
+            return None
+        spec = torch.fft.rfft(t * torch.hann_window(t.numel())).abs().pow(2) + 1e-12
+        return float(torch.exp(torch.mean(torch.log(spec))) / torch.mean(spec))
+    except Exception:  # never let the checker itself block a render
+        return None
+
+
+def _is_unusable_audio(audio_tensor) -> bool:
+    """True if a render is silent/non-finite OR a degenerate tonal buzz.
+
+    The blank guard alone misses the tonal-collapse failure mode: a buzz is
+    *loud* (peaks near -2 dBFS after normalize), so it sails past the silence
+    floor and — without this — gets cached and served as the preview.
+    """
+    if _is_blank_audio(audio_tensor):
+        return True
+    flatness = _spectral_flatness(audio_tensor)
+    return flatness is not None and flatness < _DEGENERATE_FLATNESS
+
+
 async def _render_archetype_wav(a: dict, out_path: Path) -> None:
     """Render an archetype's sample script to ``out_path`` using the live engine.
 
@@ -112,7 +158,7 @@ async def _render_archetype_wav(a: dict, out_path: Path) -> None:
             None,           # ref_text
             a["instruct"],  # instruct
             None,           # duration
-            16,             # num_step
+            _PREVIEW_NUM_STEP,  # num_step
             2.0,            # guidance_scale
             1.0,            # speed
             None,           # t_shift
@@ -126,13 +172,14 @@ async def _render_archetype_wav(a: dict, out_path: Path) -> None:
         )
 
     audio_tensor = await loop.run_in_executor(_gpu_pool, _infer, _PREVIEW_SEED)
-    if _is_blank_audio(audio_tensor):
-        # Static message only — the archetype id derives from the request path
-        # param, and CodeQL flags logging request-derived data (clear-text /
-        # log-injection). The seed is a module constant, safe to log.
-        logger.warning("Archetype rendered blank at seed %d — retrying once", _PREVIEW_SEED)
+    if _is_unusable_audio(audio_tensor):
+        # Blank OR a degenerate tonal buzz — retry once on a different seed to
+        # step off the bad diffusion trajectory. Static message only: the
+        # archetype id is request-derived (CodeQL log-injection); the seed is a
+        # module constant, safe to log.
+        logger.warning("Archetype rendered unusable at seed %d — retrying once", _PREVIEW_SEED)
         audio_tensor = await loop.run_in_executor(_gpu_pool, _infer, _PREVIEW_SEED + 1)
-    if _is_blank_audio(audio_tensor):
+    if _is_unusable_audio(audio_tensor):
         raise RuntimeError("the voice engine returned no audible audio for this archetype")
 
     out_path.parent.mkdir(parents=True, exist_ok=True)
@@ -200,7 +247,14 @@ async def preview_archetype(archetype_id: str):
                     f"unavailable. See Settings → Logs → Backend. Error: {e}"
                 ),
             )
-    return FileResponse(str(cache_path), media_type="audio/wav")
+    # no-cache (not no-store): the URL is stable but its bytes change when an
+    # archetype's preview is re-rendered, so force the client to revalidate
+    # against the ETag instead of serving a stale cached clip indefinitely.
+    return FileResponse(
+        str(cache_path),
+        media_type="audio/wav",
+        headers={"Cache-Control": "no-cache"},
+    )
 
 
 @router.post("/archetypes/{archetype_id}/use")

backend/tests/test_archetype_preview_quality.py

@@ -0,0 +1,103 @@
+"""Unit tests for the archetype-preview quality guard (``api.routers.archetypes``).
+
+Background: the Hype Host / Podcaster / Vlogger previews shipped a loud tonal
+*buzz* instead of speech. The renderer pinned ``num_step=16`` + ``seed=42`` and
+the "social" sample script collapsed to a near-pure tone at that point; the
+old silence-only guard missed it (the buzz is loud, not silent) so the garbage
+was cached and served.
+
+These tests cover the fix *without the 5 GB model / a GPU*: they drive the pure
+``_spectral_flatness`` / ``_is_unusable_audio`` helpers with synthetic signals,
+and assert the render constants didn't regress. The real end-to-end render is
+verified manually (spectral flatness back in the speech range + Whisper ASR).
+"""
+from __future__ import annotations
+
+import math
+import os
+import sys
+import tempfile
+import types
+from pathlib import Path
+
+import pytest
+
+sys.path.insert(0, os.path.dirname(os.path.dirname(__file__)))
+
+# Stub core.config before the router imports OUTPUTS_DIR / VOICES_DIR from it.
+_TMP = tempfile.mkdtemp(prefix="omnivoice_preview_q_")
+_config = types.ModuleType("core.config")
+_config.DATA_DIR = _TMP
+_config.VOICES_DIR = str(Path(_TMP) / "voices")
+_config.OUTPUTS_DIR = str(Path(_TMP) / "outputs")
+sys.modules["core.config"] = _config
+
+torch = pytest.importorskip("torch")  # noqa: E402
+
+from api.routers import archetypes as arch  # noqa: E402
+
+SR = 24_000
+N = SR * 3  # 3 s clips
+
+
+def _pure_tone(hz: float = 220.0) -> "torch.Tensor":
+    t = torch.arange(N, dtype=torch.float32) / SR
+    return 0.8 * torch.sin(2 * math.pi * hz * t)
+
+
+def _white_noise() -> "torch.Tensor":
+    g = torch.Generator().manual_seed(0)
+    return 0.5 * (torch.rand(N, generator=g) * 2 - 1)
+
+
+def _speech_like() -> "torch.Tensor":
+    """Broadband + harmonic + amplitude-modulated — a coarse stand-in for voiced
+    speech: several harmonics (formant-ish), additive noise (consonants), and a
+    syllabic envelope (word gaps). Flatness lands between a pure tone and noise.
+    """
+    g = torch.Generator().manual_seed(1)
+    t = torch.arange(N, dtype=torch.float32) / SR
+    harm = sum(torch.sin(2 * math.pi * f * t) / (i + 1)
+               for i, f in enumerate((130.0, 260.0, 390.0, 520.0)))
+    noise = 0.3 * (torch.rand(N, generator=g) * 2 - 1)
+    env = 0.5 + 0.5 * torch.sin(2 * math.pi * 4.0 * t).clamp(min=0)  # ~4 Hz syllables
+    sig = (harm + noise) * env
+    return 0.7 * sig / sig.abs().max()
+
+
+# ── _spectral_flatness ──────────────────────────────────────────────────────
+def test_flatness_orders_tone_below_speech_below_noise():
+    tone = arch._spectral_flatness(_pure_tone())
+    speech = arch._spectral_flatness(_speech_like())
+    noise = arch._spectral_flatness(_white_noise())
+    assert tone is not None and speech is not None and noise is not None
+    assert tone < arch._DEGENERATE_FLATNESS < speech < noise
+
+
+def test_flatness_returns_none_on_too_short_or_nonfinite():
+    assert arch._spectral_flatness(torch.zeros(16)) is None
+    bad = torch.full((4096,), float("nan"))
+    assert arch._spectral_flatness(bad) is None
+
+
+# ── _is_unusable_audio ──────────────────────────────────────────────────────
+def test_pure_tone_is_unusable():
+    # The degenerate-buzz failure mode: loud (passes the silence guard) but tonal.
+    tone = _pure_tone()
+    assert tone.abs().max() > 0.02            # not silent
+    assert arch._is_unusable_audio(tone) is True
+
+
+def test_silence_is_unusable():
+    assert arch._is_unusable_audio(torch.zeros(N)) is True
+
+
+def test_speech_like_is_usable():
+    assert arch._is_unusable_audio(_speech_like()) is False
+
+
+# ── Constants didn't regress ────────────────────────────────────────────────
+def test_preview_render_constants():
+    # 16 steps under-converged on the social script; the fix bumped it.
+    assert arch._PREVIEW_NUM_STEP >= 24
+    assert 0 < arch._DEGENERATE_FLATNESS < 0.03

bun.lock

@@ -54,6 +54,7 @@
       },
       "devDependencies": {
         "@eslint/js": "^10.0.1",
+        "@playwright/test": "^1.60.0",
         "@tauri-apps/api": "^2.11.0",
         "@tauri-apps/cli": "^2.11.0",
         "@testing-library/jest-dom": "^6.9.1",
@@ -203,6 +204,8 @@
 
     "@oxc-project/types": ["@oxc-project/types@0.127.0", "", {}, "sha512-aIYXQBo4lCbO4z0R3FHeucQHpF46l2LbMdxRvqvuRuW2OxdnSkcng5B8+K12spgLDj93rtN3+J2Vac/TIO+ciQ=="],
 
+    "@playwright/test": ["@playwright/test@1.60.0", "", { "dependencies": { "playwright": "1.60.0" }, "bin": { "playwright": "cli.js" } }, "sha512-O71yZIbAh/PxDMNGns37GHBIfrVkEVyn+AXyIa5dOTfb4/xNvRWV+Vv/NMbNCtODB/pO7vLlF2OTmMVLhmr7Ag=="],
+
     "@radix-ui/number": ["@radix-ui/number@1.1.1", "", {}, "sha512-MkKCwxlXTgz6CFoJx3pCwn07GKp36+aZyu/u2Ln2VrA5DcdyCZkASEDBTd8x5whTQQL5CiYf4prXKLcgQdv29g=="],
 
     "@radix-ui/primitive": ["@radix-ui/primitive@1.1.3", "", {}, "sha512-JTF99U/6XIjCBo0wqkU5sK10glYe27MRRsfwoiq5zzOEZLHU3A3KCMa5X/azekYRCJ0HlwI0crAXS/5dEHTzDg=="],

frontend/e2e/_helpers.ts

@@ -0,0 +1,49 @@
+import type { Page } from '@playwright/test';
+
+/** Every routable view (the `mode` values in App.jsx). */
+export const MODES = [
+  'launchpad', 'clone', 'design', 'gallery', 'dub', 'stories',
+  'projects', 'queue', 'tools', 'transcriptions', 'settings', 'donate',
+] as const;
+
+/**
+ * Fatal client errors that mean a view failed to LOAD — code-split chunk /
+ * dynamic-import failures (the "Use design → Importing a module script failed"
+ * regression) and uncaught exceptions. Deliberately NOT matching network/API
+ * noise (5xx, fetch failures) — backend health is covered elsewhere, and a
+ * flaky API shouldn't fail a UI-mount test.
+ */
+const FATAL = [
+  /Importing a module script failed/i,
+  /Failed to fetch dynamically imported module/i,
+  /error loading dynamically imported module/i,
+  /ChunkLoadError/i,
+];
+
+export type ErrorSink = { fatal: string[]; all: string[] };
+
+/** Attach console/pageerror listeners; returns a sink you assert on later. */
+export function collectErrors(page: Page): ErrorSink {
+  const sink: ErrorSink = { fatal: [], all: [] };
+  const record = (text: string) => {
+    sink.all.push(text);
+    if (FATAL.some((re) => re.test(text))) sink.fatal.push(text);
+  };
+  page.on('pageerror', (err) => record(`pageerror: ${err.message}`));
+  page.on('console', (msg) => {
+    if (msg.type() === 'error') record(`console.error: ${msg.text()}`);
+  });
+  return sink;
+}
+
+/**
+ * Land directly on a view by seeding the zustand-persist store (key
+ * `omnivoice.app`) before the app boots. A shallow merge over slice defaults,
+ * so only `mode` is forced.
+ */
+export async function gotoMode(page: Page, mode: string): Promise<void> {
+  await page.addInitScript((m) => {
+    localStorage.setItem('omnivoice.app', JSON.stringify({ state: { mode: m }, version: 4 }));
+  }, mode);
+  await page.goto('/');
+}

frontend/e2e/gallery.spec.ts

@@ -0,0 +1,39 @@
+import { test, expect } from '@playwright/test';
+import { collectErrors, gotoMode } from './_helpers';
+
+test.describe('OmniVoice Gallery', () => {
+  test('heading is "OmniVoice Gallery"', async ({ page }) => {
+    await gotoMode(page, 'gallery');
+    await expect(page.getByRole('heading', { name: /OmniVoice Gallery/i })).toBeVisible();
+  });
+
+  test('facet dropdowns use the dark theme, not the OS-default light surface', async ({ page }) => {
+    await gotoMode(page, 'gallery');
+    const select = page.locator('select.facet-select').first();
+    await expect(select).toBeVisible();
+    // Regression guard for the undefined-var fallback: the fixed style resolves
+    // --chrome-hover-bg → rgba(255,255,255,0.04), NOT an opaque UA light surface
+    // and NOT transparent (rgba(0,0,0,0), the broken undefined-var state).
+    const bg = await select.evaluate((el) => getComputedStyle(el).backgroundColor);
+    expect(bg).toBe('rgba(255, 255, 255, 0.04)');
+  });
+
+  test('opening an archetype in the Designer mounts the design view (no chunk-load failure)', async ({ page }) => {
+    const errors = collectErrors(page);
+    await gotoMode(page, 'gallery');
+
+    // Cards load from the backend; wait for the first one.
+    const designerBtn = page.locator('.archetype-card .designer-btn').first();
+    await expect(designerBtn).toBeVisible({ timeout: 20_000 });
+    await designerBtn.click();
+
+    // The design view (CloneDesignTab — the lazy chunk that failed when Vite
+    // was down) must mount. Its prompt/personality UI is the tell.
+    await expect(
+      page.getByText(/personality|prompt|steps/i).first()
+    ).toBeVisible({ timeout: 15_000 });
+
+    await expect(page.getByText(/this tab hit a snag/i)).toHaveCount(0);
+    expect(errors.fatal, errors.fatal.join('\n')).toEqual([]);
+  });
+});

frontend/e2e/ui-smoke.spec.ts

@@ -0,0 +1,24 @@
+import { test, expect } from '@playwright/test';
+import { MODES, collectErrors, gotoMode } from './_helpers';
+
+// Every view must mount without a code-split/import failure or an uncaught
+// exception, and without tripping the ErrorBoundary fallback. This is the
+// regression guard for "Use design → Importing a module script failed" (a dead
+// Vite/module server) and any lazy() page that fails to load.
+for (const mode of MODES) {
+  test(`view "${mode}" mounts without fatal client errors`, async ({ page }) => {
+    const errors = collectErrors(page);
+    await gotoMode(page, mode);
+
+    // Give the lazy chunk time to fetch + the Suspense boundary to resolve.
+    // (No networkidle wait — views with a live WS/SSE log stream, e.g. Settings,
+    // never reach it.)
+    await page.waitForTimeout(2000);
+
+    // The ErrorBoundary fallback copy ("This tab hit a snag.") must not show.
+    const snag = page.getByText(/this tab hit a snag/i);
+    await expect(snag).toHaveCount(0);
+
+    expect(errors.fatal, `fatal errors in "${mode}":\n${errors.fatal.join('\n')}`).toEqual([]);
+  });
+}

frontend/package.json

@@ -14,7 +14,8 @@
     "test:watch": "vitest",
     "test:legacy": "node --experimental-strip-types --no-warnings --test ../tests/frontend/*.test.mjs",
     "preview": "vite preview",
-    "tauri": "tauri"
+    "tauri": "tauri",
+    "e2e": "playwright test"
   },
   "dependencies": {
     "@fontsource-variable/inter": "^5.2.8",
@@ -54,6 +55,7 @@
   },
   "devDependencies": {
     "@eslint/js": "^10.0.1",
+    "@playwright/test": "^1.60.0",
     "@tauri-apps/api": "^2.11.0",
     "@tauri-apps/cli": "^2.11.0",
     "@testing-library/jest-dom": "^6.9.1",

frontend/playwright.config.ts

@@ -0,0 +1,31 @@
+import { defineConfig, devices } from '@playwright/test';
+
+// E2E runs against the Vite dev server (UI on :3901, backend on :3900). It
+// drives the SYSTEM chromium (no `playwright install` browser download) — set
+// PLAYWRIGHT_CHROMIUM to override the path. reuseExistingServer keeps a dev
+// session you already have running; CI starts its own `bun run dev`.
+const PORT = Number(process.env.E2E_PORT || 3901);
+
+export default defineConfig({
+  testDir: './e2e',
+  timeout: 45_000,
+  expect: { timeout: 10_000 },
+  fullyParallel: false,
+  retries: process.env.CI ? 1 : 0,
+  reporter: [['list']],
+  use: {
+    baseURL: `http://localhost:${PORT}`,
+    headless: true,
+    trace: 'retain-on-failure',
+    launchOptions: {
+      executablePath: process.env.PLAYWRIGHT_CHROMIUM || '/usr/bin/chromium',
+    },
+  },
+  projects: [{ name: 'chromium', use: { ...devices['Desktop Chrome'] } }],
+  webServer: {
+    command: 'bun run dev',
+    url: `http://localhost:${PORT}`,
+    reuseExistingServer: true,
+    timeout: 60_000,
+  },
+});

frontend/src/i18n/locales/ar.json

@@ -851,7 +851,7 @@
     "add_to_queue": "إضافة إلى قائمة الانتظار"
   },
   "gallery": {
-    "title": "معرض",
+    "title": "OmniVoice معرض",
     "search_placeholder": "بحث في يوتيوب...",
     "all_voices": "جميع الأصوات ({{count}})",
     "no_voices": "لا توجد أصوات بعد",

frontend/src/i18n/locales/de.json

@@ -852,7 +852,7 @@
     "add_to_queue": "Zur Warteschlange hinzufügen"
   },
   "gallery": {
-    "title": "Galerie",
+    "title": "OmniVoice Galerie",
     "search_placeholder": "YouTube durchsuchen…",
     "all_voices": "Alle Stimmen ({{count}})",
     "no_voices": "Noch keine Stimmen",

frontend/src/i18n/locales/en.json

@@ -730,7 +730,7 @@
     "add_to_queue": "Add to Queue"
   },
   "gallery": {
-    "title": "Gallery",
+    "title": "OmniVoice Gallery",
     "search_placeholder": "Search YouTube…",
     "all_voices": "All Voices ({{count}})",
     "no_voices": "No voices yet",

frontend/src/i18n/locales/es.json

@@ -852,7 +852,7 @@
     "add_to_queue": "Agregar a la cola"
   },
   "gallery": {
-    "title": "Galería",
+    "title": "OmniVoice Galería",
     "search_placeholder": "Buscar en YouTube…",
     "all_voices": "Todas las voces ({{count}})",
     "no_voices": "Aún no hay voces",

frontend/src/i18n/locales/fr.json

@@ -852,7 +852,7 @@
     "add_to_queue": "Ajouter à la file d'attente"
   },
   "gallery": {
-    "title": "Galerie",
+    "title": "OmniVoice Galerie",
     "search_placeholder": "Rechercher sur YouTube…",
     "all_voices": "Toutes les voix ({{count}})",
     "no_voices": "Pas encore de voix",

frontend/src/i18n/locales/hi.json

@@ -851,7 +851,7 @@
     "add_to_queue": "कतार में जोड़ें"
   },
   "gallery": {
-    "title": "गैलरी",
+    "title": "OmniVoice गैलरी",
     "search_placeholder": "यूट्यूब पर खोजें...",
     "all_voices": "सभी आवाज़ें ({{count}})",
     "no_voices": "अभी तक कोई आवाज़ नहीं",

frontend/src/i18n/locales/id.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Tambahkan ke Antrean"
   },
   "gallery": {
-    "title": "Galeri",
+    "title": "OmniVoice Galeri",
     "search_placeholder": "Telusuri YouTube…",
     "all_voices": "Semua Suara ({{count}})",
     "no_voices": "Belum ada suara",

frontend/src/i18n/locales/it.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Aggiungi alla coda"
   },
   "gallery": {
-    "title": "Galleria",
+    "title": "OmniVoice Galleria",
     "search_placeholder": "Cerca su YouTube...",
     "all_voices": "Tutte le voci ({{count}})",
     "no_voices": "Nessuna voce ancora",

frontend/src/i18n/locales/ja.json

@@ -852,7 +852,7 @@
     "add_to_queue": "キューに追加"
   },
   "gallery": {
-    "title": "ギャラリー",
+    "title": "OmniVoice ギャラリー",
     "search_placeholder": "YouTube を検索…",
     "all_voices": "すべての声 ({{count}})",
     "no_voices": "まだ声はありません",

frontend/src/i18n/locales/ko.json

@@ -851,7 +851,7 @@
     "add_to_queue": "대기열에 추가"
   },
   "gallery": {
-    "title": "갤러리",
+    "title": "OmniVoice 갤러리",
     "search_placeholder": "유튜브 검색…",
     "all_voices": "모든 음색({{count}})",
     "no_voices": "아직 음성이 없습니다.",

frontend/src/i18n/locales/nl.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Toevoegen aan wachtrij"
   },
   "gallery": {
-    "title": "Galerij",
+    "title": "OmniVoice Galerij",
     "search_placeholder": "Zoek op YouTube...",
     "all_voices": "Alle stemmen ({{count}})",
     "no_voices": "Nog geen stemmen",

frontend/src/i18n/locales/pl.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Dodaj do kolejki"
   },
   "gallery": {
-    "title": "Galeria",
+    "title": "OmniVoice Galeria",
     "search_placeholder": "Wyszukaj w YouTube…",
     "all_voices": "Wszystkie głosy ({{count}})",
     "no_voices": "Nie ma jeszcze głosów",

frontend/src/i18n/locales/pt.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Adicionar à fila"
   },
   "gallery": {
-    "title": "Galeria",
+    "title": "OmniVoice Galeria",
     "search_placeholder": "Pesquisar no YouTube…",
     "all_voices": "Todas as vozes ({{count}})",
     "no_voices": "Ainda não há vozes",

frontend/src/i18n/locales/ru.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Добавить в очередь"
   },
   "gallery": {
-    "title": "Галерея",
+    "title": "OmniVoice Галерея",
     "search_placeholder": "Искать на YouTube…",
     "all_voices": "Все голоса ({{count}})",
     "no_voices": "Голосов пока нет",

frontend/src/i18n/locales/sv.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Lägg till i kö"
   },
   "gallery": {
-    "title": "Galleri",
+    "title": "OmniVoice Galleri",
     "search_placeholder": "Sök på YouTube...",
     "all_voices": "Alla röster ({{count}})",
     "no_voices": "Inga röster än",

frontend/src/i18n/locales/th.json

@@ -851,7 +851,7 @@
     "add_to_queue": "เพิ่มเข้าคิว"
   },
   "gallery": {
-    "title": "แกลเลอรี่",
+    "title": "OmniVoice แกลเลอรี่",
     "search_placeholder": "ค้นหา YouTube...",
     "all_voices": "ทุกเสียง ({{count}})",
     "no_voices": "ยังไม่มีเสียง.",

frontend/src/i18n/locales/tr.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Kuyruğa Ekle"
   },
   "gallery": {
-    "title": "Galeri",
+    "title": "OmniVoice Galeri",
     "search_placeholder": "YouTube'da ara…",
     "all_voices": "Tüm Sesler ({{count}})",
     "no_voices": "Henüz ses yok",

frontend/src/i18n/locales/uk.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Додати в чергу"
   },
   "gallery": {
-    "title": "Галерея",
+    "title": "OmniVoice Галерея",
     "search_placeholder": "Пошук на YouTube…",
     "all_voices": "Усі голоси ({{count}})",
     "no_voices": "Голосів ще немає",

frontend/src/i18n/locales/vi.json

@@ -851,7 +851,7 @@
     "add_to_queue": "Thêm vào hàng đợi"
   },
   "gallery": {
-    "title": "Thư viện ảnh",
+    "title": "OmniVoice Thư viện ảnh",
     "search_placeholder": "Tìm kiếm YouTube…",
     "all_voices": "Tất cả các giọng nói ({{count}})",
     "no_voices": "Chưa có tiếng nói nào",

frontend/src/i18n/locales/zh-CN.json

@@ -821,7 +821,7 @@
     "add_to_queue": "添加到队列"
   },
   "gallery": {
-    "title": "音色库",
+    "title": "OmniVoice 音色库",
     "search_placeholder": "搜索 YouTube…",
     "all_voices": "全部声音（{{count}}）",
     "no_voices": "暂无声音",

frontend/src/i18n/locales/zh-TW.json

@@ -851,7 +851,7 @@
     "add_to_queue": "添加到隊列"
   },
   "gallery": {
-    "title": "畫廊",
+    "title": "OmniVoice 畫廊",
     "search_placeholder": "搜尋 YouTube...",
     "all_voices": "所有聲音 ({{count}})",
     "no_voices": "還沒有聲音",

frontend/src/pages/VoiceGallery.css

@@ -363,7 +363,14 @@
 .use-case-chips { display: flex; flex-wrap: wrap; gap: 5px; }
 .chip-emoji { font-size: 0.8rem; line-height: 1; }
 .facet-selects { display: flex; flex-wrap: wrap; align-items: center; gap: 6px; }
-.facet-select { padding: 5px 8px; border-radius: 8px; border: 1px solid var(--border-color); background: var(--bg-tertiary); color: var(--text-primary); font-size: 0.7rem; cursor: pointer; }
+/* Native <select> falls back to the OS-default (light) control surface when its
+   background resolves to an undefined var, so pin real dark-chrome tokens. The
+   undefined --border-color/--bg-tertiary read as transparent on the sibling
+   <div> filters (fine on the dark page) but render light on a form control. */
+.facet-select { padding: 5px 8px; border-radius: 8px; border: 1px solid var(--chrome-border); background: var(--chrome-hover-bg); color: var(--chrome-fg); font-size: 0.7rem; cursor: pointer; color-scheme: dark; }
+/* Belt-and-suspenders: explicit dark option list for WebViews that ignore
+   color-scheme on the popup (cross-platform parity: WebKit / WebView2 / WebKitGTK). */
+.facet-select option { background: var(--chrome-bg); color: var(--chrome-fg); }
 .facet-toggle { display: inline-flex; align-items: center; gap: 4px; font-size: 0.7rem; color: var(--text-secondary); cursor: pointer; }
 .facet-reset { display: inline-flex; align-items: center; gap: 4px; padding: 5px 8px; border: 1px solid var(--border-color); background: transparent; color: var(--text-secondary); border-radius: 8px; font-size: 0.7rem; cursor: pointer; }
 .facet-reset:hover { border-color: var(--border-hover); color: var(--text-primary); }

frontend/src/pages/VoiceGallery.jsx

@@ -97,7 +97,10 @@ export default function VoiceGallery() {
     stopPlayback();
     setLoadingPreviewId(id);
     try {
-      const resp = await fetch(fullUrl);
+      // no-store: preview audio is re-rendered server-side when an archetype is
+      // fixed/changed, but the URL is stable. Without this the WebView's HTTP
+      // cache replays the first clip it ever fetched (a stale render) forever.
+      const resp = await fetch(fullUrl, { cache: 'no-store' });
       if (!resp.ok) throw new Error(`HTTP ${resp.status}`);
       const blob = await resp.blob();
       setPlayingId(id);
@@ -131,7 +134,7 @@ export default function VoiceGallery() {
       <div className="gallery-header">
         <div className="header-top">
           <div className="header-text">
-            <h2>{t('gallery.title', { defaultValue: 'Voice Gallery' })}</h2>
+            <h2>{t('gallery.title', { defaultValue: 'OmniVoice Gallery' })}</h2>
             <p className="gallery-sub">
               {t('gallery.subtitle', { defaultValue: 'Hundreds of ready-made designed voices — pick one and go.' })}
             </p>