Fix build error

app.py

@@ -1,4 +1,5 @@
 import os
+import json
 import shutil
 import socket
 from pathlib import Path
@@ -23,6 +24,8 @@ from backend.omnivoice_adapter import OmniVoiceAdapter
 from backend.render_pipeline import RenderPipeline
 from backend.session_store import SessionStore
 from backend.types import VoiceConfig
+from backend.voice_design import VOICE_DESIGN_OPTIONS
+from backend.voice_presets import VOICE_PRESETS
 
 
 ROOT = Path(__file__).parent
@@ -197,7 +200,14 @@ def export_audiobook_api(
 
 @app.get("/", response_class=HTMLResponse)
 async def homepage() -> HTMLResponse:
-    return HTMLResponse((FRONTEND_DIR / "index.html").read_text(encoding="utf-8"))
+    html = (FRONTEND_DIR / "index.html").read_text(encoding="utf-8")
+    preset_script = (
+        "<script>"
+        f"window.__VOICE_PRESETS__ = {json.dumps(VOICE_PRESETS)};"
+        f"window.__VOICE_DESIGN_OPTIONS__ = {json.dumps(VOICE_DESIGN_OPTIONS)};"
+        "</script>"
+    )
+    return HTMLResponse(html.replace("</body>", f"    {preset_script}\n  </body>"))
 
 
 @app.get("/health")

backend/omnivoice_adapter.py

@@ -6,6 +6,7 @@ import numpy as np
 import soundfile as sf
 
 from backend.types import VoiceConfig
+from backend.voice_presets import VOICE_PRESET_BY_ID
 
 try:
     import spaces
@@ -22,15 +23,6 @@ except ImportError:
 
     spaces = _SpacesShim()
 
-
-NARRATOR_PRESETS = {
-    "brother-anselm": "male, warm baritone, measured, british accent",
-    "dame-eglantine": "female, bright contralto, lyrical, british accent",
-    "the-reeve": "male, gravelly bass, wry, british accent",
-    "sister-cecilia": "female, clear soprano, gentle, british accent",
-}
-
-
 class OmniVoiceAdapter:
     def __init__(
         self,
@@ -99,7 +91,9 @@ class OmniVoiceAdapter:
         elif voice_config.mode == "design":
             kwargs["instruct"] = voice_config.design_prompt
         elif voice_config.narrator_id:
-            kwargs["instruct"] = NARRATOR_PRESETS.get(voice_config.narrator_id)
+            preset = VOICE_PRESET_BY_ID.get(voice_config.narrator_id)
+            if preset:
+                kwargs["instruct"] = preset["instruct"]
 
         audio = model.generate(**kwargs)
         waveform = np.asarray(audio[0], dtype=np.float32)
@@ -129,7 +123,9 @@ class OmniVoiceAdapter:
         elif voice_config.mode == "clone":
             base_freq = 140.0
         elif voice_config.narrator_id:
-            base_freq = 160.0 + (list(NARRATOR_PRESETS.keys()).index(voice_config.narrator_id) * 20)
+            preset_ids = list(VOICE_PRESET_BY_ID.keys())
+            if voice_config.narrator_id in preset_ids:
+                base_freq = 160.0 + (preset_ids.index(voice_config.narrator_id) * 20)
 
         timeline = np.linspace(0, duration_seconds, total_samples, endpoint=False)
         waveform = (

backend/voice_design.py

@@ -0,0 +1,55 @@
+VOICE_DESIGN_OPTIONS = [
+    {
+        "key": "gender",
+        "label": "Gender",
+        "options": [
+            {"value": "male", "label": "Male"},
+            {"value": "female", "label": "Female"},
+        ],
+    },
+    {
+        "key": "age",
+        "label": "Age",
+        "options": [
+            {"value": "child", "label": "Child"},
+            {"value": "teenager", "label": "Teenager"},
+            {"value": "young adult", "label": "Young Adult"},
+            {"value": "middle-aged", "label": "Middle-aged"},
+            {"value": "elderly", "label": "Elderly"},
+        ],
+    },
+    {
+        "key": "pitch",
+        "label": "Pitch",
+        "options": [
+            {"value": "very low pitch", "label": "Very Low Pitch"},
+            {"value": "low pitch", "label": "Low Pitch"},
+            {"value": "moderate pitch", "label": "Moderate Pitch"},
+            {"value": "high pitch", "label": "High Pitch"},
+            {"value": "very high pitch", "label": "Very High Pitch"},
+        ],
+    },
+    {
+        "key": "style",
+        "label": "Style",
+        "options": [
+            {"value": "whisper", "label": "Whisper"},
+        ],
+    },
+    {
+        "key": "accent",
+        "label": "Accent",
+        "options": [
+            {"value": "american accent", "label": "American Accent"},
+            {"value": "australian accent", "label": "Australian Accent"},
+            {"value": "british accent", "label": "British Accent"},
+            {"value": "canadian accent", "label": "Canadian Accent"},
+            {"value": "chinese accent", "label": "Chinese Accent"},
+            {"value": "indian accent", "label": "Indian Accent"},
+            {"value": "japanese accent", "label": "Japanese Accent"},
+            {"value": "korean accent", "label": "Korean Accent"},
+            {"value": "portuguese accent", "label": "Portuguese Accent"},
+            {"value": "russian accent", "label": "Russian Accent"},
+        ],
+    },
+]

backend/voice_presets.py

@@ -0,0 +1,32 @@
+VOICE_PRESETS = [
+    {
+        "id": "british-male",
+        "name": "British Male",
+        "desc": "Voice design preset · male · low pitch · british accent",
+        "initial": "B",
+        "instruct": "male, low pitch, british accent",
+    },
+    {
+        "id": "british-female",
+        "name": "British Female",
+        "desc": "Voice design preset · female · moderate pitch · british accent",
+        "initial": "B",
+        "instruct": "female, moderate pitch, british accent",
+    },
+    {
+        "id": "american-male",
+        "name": "American Male",
+        "desc": "Voice design preset · male · middle-aged · american accent",
+        "initial": "A",
+        "instruct": "male, middle-aged, american accent",
+    },
+    {
+        "id": "canadian-female",
+        "name": "Canadian Female",
+        "desc": "Voice design preset · female · young adult · canadian accent",
+        "initial": "C",
+        "instruct": "female, young adult, canadian accent",
+    },
+]
+
+VOICE_PRESET_BY_ID = {preset["id"]: preset for preset in VOICE_PRESETS}

frontend/app.js

@@ -1,11 +1,13 @@
 import { Client, handle_file } from "https://esm.sh/@gradio/client";
 
-const NARRATORS = [
-  { id: "brother-anselm", name: "The Archivist", desc: "OmniVoice preset · warm baritone · measured", initial: "A" },
-  { id: "dame-eglantine", name: "Lady Meridian", desc: "OmniVoice preset · bright contralto · lyrical", initial: "M" },
-  { id: "the-reeve", name: "North Bell", desc: "OmniVoice preset · gravelled bass · wry", initial: "N" },
-  { id: "sister-cecilia", name: "Cecilia of the Lamp", desc: "OmniVoice preset · clear soprano · gentle", initial: "C" },
-];
+const NARRATORS = globalThis.__VOICE_PRESETS__ || [];
+const VOICE_DESIGN_OPTIONS = globalThis.__VOICE_DESIGN_OPTIONS__ || [];
+const DEFAULT_NARRATOR = NARRATORS[0] || {
+  id: "british-male",
+  name: "British Male",
+  desc: "Voice design preset · male · low pitch · british accent",
+  initial: "B",
+};
 
 const VOICE_WARNING =
   "Use clone mode only with audio you have the right to use. Unauthorized voice cloning, impersonation, fraud, or scams are prohibited.";
@@ -18,11 +20,17 @@ const state = {
   chapters: [],
   currentChapterId: null,
   voiceMode: "auto",
-  narratorId: "brother-anselm",
+  narratorId: DEFAULT_NARRATOR.id,
   cloneConsent: false,
   cloneSampleFile: null,
   cloneReferenceText: "",
-  designPrompt: "",
+  designSelections: {
+    gender: "",
+    age: "",
+    pitch: "",
+    style: "",
+    accent: "",
+  },
   diffusionSteps: 32,
   speed: 1,
   statusMessage: "",
@@ -38,7 +46,7 @@ const state = {
   exportMetadata: {
     title: "",
     author: "",
-    narrator: "Brother Anselm (OmniVoice)",
+    narrator: `${DEFAULT_NARRATOR.name} (OmniVoice)`,
     genre: "Audiobook",
   },
 };
@@ -106,6 +114,11 @@ function handleChange(event) {
   if (target.matches("#embed-markers")) {
     state.embedMarkers = target.checked;
     render();
+    return;
+  }
+  if (target.matches("[data-design-field]")) {
+    state.designSelections[target.dataset.designField] = target.value;
+    render();
   }
 }
 
@@ -125,10 +138,6 @@ function handleInput(event) {
     state.cloneReferenceText = target.value;
     return;
   }
-  if (target.matches("#design-prompt")) {
-    state.designPrompt = target.value;
-    return;
-  }
   if (target.matches("#clone-consent")) {
     state.cloneConsent = target.checked;
     render();
@@ -378,7 +387,7 @@ function buildVoiceConfig() {
   if (state.voiceMode === "design") {
     return {
       mode: "design",
-      designPrompt: state.designPrompt,
+      designPrompt: buildDesignPrompt(),
     };
   }
   return {
@@ -868,11 +877,29 @@ function renderVoiceMode() {
     `;
   }
   if (state.voiceMode === "design") {
+    const prompt = buildDesignPrompt();
     return `
-      <label style="display:block;">
-        <div class="smallcaps" style="margin-bottom:7px;">Describe the voice you want</div>
-        <textarea id="design-prompt" class="textarea" placeholder="A warm, unhurried narrator — low pitch, faintly British, the kind of voice that belongs by a fireplace…">${escapeHtml(state.designPrompt)}</textarea>
-      </label>
+      <div class="smallcaps" style="margin-bottom:10px;">Build a supported OmniVoice design prompt</div>
+      <div class="design-grid">
+        ${VOICE_DESIGN_OPTIONS.map((category) => `
+          <label style="display:block;">
+            <div class="smallcaps" style="margin-bottom:7px;">${escapeHtml(category.label)}</div>
+            <select class="field" data-design-field="${escapeAttr(category.key)}">
+              <option value="">None</option>
+              ${category.options.map((option) => `
+                <option value="${escapeAttr(option.value)}" ${state.designSelections[category.key] === option.value ? "selected" : ""}>
+                  ${escapeHtml(option.label)}
+                </option>
+              `).join("")}
+            </select>
+          </label>
+        `).join("")}
+      </div>
+      <div class="design-preview">
+        <div class="smallcaps" style="margin-bottom:7px;">Resulting prompt</div>
+        <div class="field design-output">${escapeHtml(prompt || "Choose one or more supported voice traits.")}</div>
+        <div class="chapter-meta" style="margin-top:8px;">Use English-only OmniVoice design tokens. Accent affects English speech only.</div>
+      </div>
     `;
   }
   return `
@@ -939,10 +966,22 @@ function canStartRender() {
   if (!state.book) return false;
   if (!selectedChapterIds().length) return false;
   if (state.voiceMode === "clone" && (!state.cloneSampleFile || !state.cloneConsent)) return false;
-  if (state.voiceMode === "design" && !state.designPrompt.trim()) return false;
+  if (state.voiceMode === "design" && !buildDesignPrompt()) return false;
   return true;
 }
 
+function buildDesignPrompt() {
+  return [
+    state.designSelections.gender,
+    state.designSelections.age,
+    state.designSelections.pitch,
+    state.designSelections.style,
+    state.designSelections.accent,
+  ]
+    .filter(Boolean)
+    .join(", ");
+}
+
 function activeNarratorName() {
   const narrator = NARRATORS.find((item) => item.id === state.narratorId);
   return narrator?.name || "Custom narrator";

frontend/styles.css

@@ -645,6 +645,22 @@ select {
   resize: vertical;
 }
 
+.design-grid {
+  display: grid;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 14px;
+}
+
+.design-preview {
+  margin-top: 16px;
+}
+
+.design-output {
+  min-height: 48px;
+  display: flex;
+  align-items: center;
+}
+
 .dial-grid {
   display: grid;
   grid-template-columns: repeat(2, minmax(0, 1fr));

tests/test_voice_presets.py

@@ -0,0 +1,39 @@
+import asyncio
+import importlib
+import sys
+
+from omnivoice.utils.voice_design import _INSTRUCT_VALID_EN
+
+from backend.voice_design import VOICE_DESIGN_OPTIONS
+from backend.voice_presets import VOICE_PRESETS
+
+
+def test_voice_presets_use_only_supported_omnivoice_design_tokens() -> None:
+    for preset in VOICE_PRESETS:
+        instruct = preset["instruct"]
+        parts = [part.strip().lower() for part in instruct.split(",")]
+        assert all(parts), f"Preset {preset['id']} contains an empty instruct token"
+        assert all(
+            part in _INSTRUCT_VALID_EN for part in parts
+        ), f"Preset {preset['id']} has unsupported instruct tokens: {instruct}"
+
+
+def test_homepage_injects_voice_presets_for_frontend() -> None:
+    sys.modules.pop("app", None)
+    module = importlib.import_module("app")
+
+    response = asyncio.run(module.homepage())
+    html = response.body.decode("utf-8")
+
+    assert "window.__VOICE_PRESETS__" in html
+    assert "window.__VOICE_DESIGN_OPTIONS__" in html
+
+
+def test_voice_design_options_use_only_supported_omnivoice_design_tokens() -> None:
+    option_values = [
+        option["value"]
+        for category in VOICE_DESIGN_OPTIONS
+        for option in category["options"]
+    ]
+    assert option_values
+    assert all(value in _INSTRUCT_VALID_EN for value in option_values)