Merge branch 'main' of https://huggingface.co/spaces/build-small-hackathon/Hackathon-IA-VisualNovel

.gitignore

@@ -1,6 +1,6 @@
 # Secrets — never commit
-.env
 .env.local
+.env.secret
 
 # Python
 __pycache__/
@@ -29,3 +29,5 @@ memory/
 # macOS
 .DS_Store
 .python-version
+
+uv.lock

app.py

@@ -104,12 +104,19 @@ if not config.USE_MOCK:
         except ImportError:
             _missing.append("transformers")
 
-    if config.IMAGE_BACKEND == "local":
+    if config.IMAGE_BACKEND in ("local", "lightning"):
         try:
             import diffusers  # noqa: F401
         except ImportError:
             _missing.append("image")
 
+    if config.TTS_BACKEND == "kokoro":
+        try:
+            import kokoro_onnx  # noqa: F401
+            import soundfile  # noqa: F401
+        except ImportError:
+            _missing.append("tts")
+
     if config.LLM_BACKEND == "modal" or config.IMAGE_BACKEND == "modal":
         try:
             import modal  # noqa: F401

frontend/index.html

@@ -143,6 +143,9 @@
   @keyframes pulse-rec{0%,100%{opacity:1}50%{opacity:.6}}
   #mic.live{background:var(--rose); animation:pulse-rec 1.2s ease-in-out infinite}
   #mic.waiting{background:rgba(255,255,255,.12); color:var(--ink-dim); cursor:wait}
+  #tts-toggle{min-width:42px; background:transparent; color:var(--ink-dim);
+    border:1px solid rgba(255,255,255,.22); font-size:1.1rem}
+  #tts-toggle.active{color:var(--gold); border-color:rgba(232,193,114,.5)}
   /* ---- book intro overlay ---- */
   #intro-book{
     position:fixed; inset:0; z-index:30; display:flex; align-items:center; justify-content:center;
@@ -354,6 +357,7 @@
         <input id="say" placeholder="say something…" autocomplete="off" />
         <button id="send">✓</button>
         <button id="mic" class="ghost" title="hold to speak">🎙️</button>
+        <button id="tts-toggle" title="toggle voice output">🔇</button>
       </div>
     </div>
   </div>
@@ -567,6 +571,7 @@ function render(v){
   $("place").textContent = v.place || "";
   $("speaker").textContent = v.speaker || "";
   typeOut($("line"), v.dialogue || "");
+  playTTS(v.audio_b64);
   if (v.backdrop_url && v.backdrop_url !== lastBg){
     const bg = $("backdrop");
     bg.classList.remove("on");
@@ -725,6 +730,30 @@ function busy(state){
   document.querySelectorAll(".target-chip").forEach(b => b.disabled = state);
 }
 
+// --- TTS toggle ---
+let ttsEnabled = true;
+const ttsBtn = $("tts-toggle");
+ttsBtn.textContent = "🔊";
+ttsBtn.classList.add("active");
+ttsBtn.addEventListener("click", () => {
+  ttsEnabled = !ttsEnabled;
+  ttsBtn.textContent = ttsEnabled ? "🔊" : "🔇";
+  ttsBtn.classList.toggle("active", ttsEnabled);
+  if (!ttsEnabled) { const a = $("vn-audio"); if (a) { a.pause(); a.src = ""; } }
+});
+
+function playTTS(audio_b64) {
+  if (!ttsEnabled || !audio_b64) return;
+  let audio = $("vn-audio");
+  if (!audio) {
+    audio = document.createElement("audio");
+    audio.id = "vn-audio";
+    document.body.appendChild(audio);
+  }
+  audio.src = audio_b64;
+  audio.play().catch(() => {});  // swallow autoplay policy errors silently
+}
+
 // --- book overlay dismiss ---
 $("book-close").addEventListener("click", () => {
   $("intro-book").classList.add("hidden");
@@ -797,10 +826,10 @@ $("mic").addEventListener("click", async () => {
       try {
         const file = new File([blob], "speech.webm", { type: "audio/webm" });
         const { data } = await client.predict("/transcribe", { audio: handle_file(file) });
-        micIdle();
-        if (data[0]?.text) { $("say").value = data[0].text; sendTurn(); return; }
+        micIdle(); mic.disabled = false; $("say").disabled = false; $("send").disabled = false;
+        if (data[0]?.text) { $("say").value = data[0].text; $("say").focus(); return; }
       } catch (e) { console.error("[stt] transcribe error", e); }
-      // fallback: restore input if no text or error
+      // fallback: restore input on error
       mic.disabled = false; $("say").disabled = false; $("send").disabled = false;
     };
     rec.start(100);

pyproject.toml

@@ -27,11 +27,13 @@ llamacpp = ["llama-cpp-python>=0.3.0"]
 transformers = ["transformers>=4.44", "accelerate>=0.33", "torch>=2.3"]
 image = ["diffusers>=0.38.0", "transformers>=4.44", "accelerate>=0.33", "torch>=2.3", "safetensors>=0.4", "rembg>=2.0", "peft>=0.11"]
 stt = ["faster-whisper>=1.0", "sounddevice>=0.5.5"]
+tts = ["kokoro-onnx>=0.3", "soundfile>=0.12"]
 matting = ["transformers>=4.44", "torch>=2.3"]   # BiRefNet for transparent sprites (Phase 2)
 spaces = ["spaces>=0.30"]                         # ZeroGPU decorator on HF Spaces
 modal = ["modal>=1.0"]                            # cloud GPU via Modal (LLM + painter)
 dev = ["ruff>=0.6", "pytest>=8.0", "matplotlib>=3.9"]
 
+
 [tool.ruff]
 line-length = 100
 target-version = "py311"

uv.lock

@@ -294,6 +294,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/5f/af/0a6e1d2a845988039f6c197fa7269b5e9abbe17354fb41cc9d75bb260fcb/av-17.0.1-cp314-cp314t-win_arm64.whl", hash = "sha256:a87a42c36e29f75e7dff7281944f2a6876a2c8875e225ccbf6c1ae62748b4caa", size = 22072676, upload-time = "2026-04-18T17:12:31.836Z" },
 ]
 
+[[package]]
+name = "babel"
+version = "2.18.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/7d/b2/51899539b6ceeeb420d40ed3cd4b7a40519404f9baf3d4ac99dc413a834b/babel-2.18.0.tar.gz", hash = "sha256:b80b99a14bd085fcacfa15c9165f651fbb3406e66cc603abf11c5750937c992d", size = 9959554, upload-time = "2026-02-01T12:30:56.078Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/77/f5/21d2de20e8b8b0408f0681956ca2c69f1320a3848ac50e6e7f39c6159675/babel-2.18.0-py3-none-any.whl", hash = "sha256:e2b422b277c2b9a9630c1d7903c2a00d0830c409c59ac8cae9081c92f1aeba35", size = 10196845, upload-time = "2026-02-01T12:30:53.445Z" },
+]
+
 [[package]]
 name = "brotli"
 version = "1.2.0"
@@ -661,6 +670,26 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/0c/58/bd257695f39d05594ca4ad60df5bcb7e32247f9951fd09a9b8edb82d1daa/contourpy-1.3.3-pp311-pypy311_pp73-win_amd64.whl", hash = "sha256:3d1a3799d62d45c18bafd41c5fa05120b96a28079f2393af559b843d1a966a77", size = 225315, upload-time = "2025-07-26T12:02:58.801Z" },
 ]
 
+[[package]]
+name = "csvw"
+version = "4.0.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "babel" },
+    { name = "isodate" },
+    { name = "jsonschema" },
+    { name = "language-tags" },
+    { name = "python-dateutil" },
+    { name = "rdflib" },
+    { name = "rfc3986" },
+    { name = "termcolor" },
+    { name = "uritemplate" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/33/50/ff0bd8c3663dad0dae929775a6078c5bb7f4038d93032369786b1fe6a91c/csvw-4.0.0.tar.gz", hash = "sha256:060e9bedf3c274d0fce6d6f7f892a16cc8c72f39c2bc49b69b5f0858fb2f6217", size = 83025, upload-time = "2026-05-05T06:25:26.787Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/3a/88/9713d1ecac111742d60e1d9c2c15fec56fd121940f97a73d014dc9a7d521/csvw-4.0.0-py2.py3-none-any.whl", hash = "sha256:df875fcb1505afd15061b5f370268522bf162640de0662a724453dcb4db6a88b", size = 69424, upload-time = "2026-05-05T06:25:24.646Z" },
+]
+
 [[package]]
 name = "ctranslate2"
 version = "4.7.2"
@@ -804,6 +833,28 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/3f/27/4570e78fc0bf5ea0ca45eb1de3818a23787af9b390c0b0a0033a1b8236f9/diskcache-5.6.3-py3-none-any.whl", hash = "sha256:5e31b2d5fbad117cc363ebaf6b689474db18a1f6438bc82358b024abd4c2ca19", size = 45550, upload-time = "2023-08-31T06:11:58.822Z" },
 ]
 
+[[package]]
+name = "dlinfo"
+version = "2.0.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/85/8e/8f2f94cd40af1b51e8e371a83b385d622170d42f98776441a6118f4dd682/dlinfo-2.0.0.tar.gz", hash = "sha256:88a2bc04f51d01bc604cdc9eb1c3cc0bde89057532ca6a3e71a41f6235433e17", size = 12727, upload-time = "2025-01-16T15:43:10.756Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/da/90/022c79d6e5e6f843268c10b84d4a021ee3afba0621d3c176d3ff2024bfc8/dlinfo-2.0.0-py3-none-any.whl", hash = "sha256:b32cc18e3ea67c0ca9ca409e5b41eed863bd1363dbc9dd3de90fedf11b61e7bc", size = 3654, upload-time = "2025-01-16T15:43:09.474Z" },
+]
+
+[[package]]
+name = "espeakng-loader"
+version = "0.2.4"
+source = { registry = "https://pypi.org/simple" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/f8/92/f44ed7f531143c3c6c97d56e2b0f9be8728dc05e18b96d46eb539230ed46/espeakng_loader-0.2.4-py3-none-macosx_10_12_x86_64.whl", hash = "sha256:b77477ae2ddf62a748e04e49714eabb2f3a24f344166200b00539083bd669904", size = 9938387, upload-time = "2025-01-17T01:22:42.064Z" },
+    { url = "https://files.pythonhosted.org/packages/a8/26/258c0cd43b9bc1043301c5f61767d6a6c3b679df82790c9cb43a3277b865/espeakng_loader-0.2.4-py3-none-macosx_11_0_arm64.whl", hash = "sha256:d27cdca31112226e7299d8562e889d3e38a1e48055c9ee381b45d669072ee59f", size = 9892565, upload-time = "2025-01-17T01:22:40.365Z" },
+    { url = "https://files.pythonhosted.org/packages/de/1e/25ec5ab07528c0fbb215a61800a38eca05c8a99445515a02d7fa5debcb32/espeakng_loader-0.2.4-py3-none-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:08721baf27d13d461f6be6eed9a65277e70d68234ff484fd8b9897b222cdcb6d", size = 10078484, upload-time = "2025-01-17T01:22:43.373Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/ad/1b768d8daffc2996e07bbcb6f534d8de3202cd75fce1f1c45eced1ce6465/espeakng_loader-0.2.4-py3-none-manylinux_2_28_aarch64.whl", hash = "sha256:d1e798141b46a050cdb75fcf3c17db969bb2c40394f3f4a48910655d547508b9", size = 10037736, upload-time = "2025-01-17T01:22:42.576Z" },
+    { url = "https://files.pythonhosted.org/packages/9d/ed/a3d872fbad4f3a3f3db0e8c31768ab14e77cd77306de16b8b20b1e1df7ea/espeakng_loader-0.2.4-py3-none-win_amd64.whl", hash = "sha256:41f1e08ac9deda2efd1ea9de0b81dab9f5ae3c4b24284f76533d0a7b1dd7abd7", size = 9437292, upload-time = "2025-01-17T01:23:27.463Z" },
+    { url = "https://files.pythonhosted.org/packages/29/64/0b75bc50ec53b4e000bac913625511215aa96124adf5dba8c4baa17c02cd/espeakng_loader-0.2.4-py3-none-win_arm64.whl", hash = "sha256:d7a2928843eaeb2df82f99a370f44e8a630f59b02f9b0d1f168a03c4eeb76b89", size = 9426841, upload-time = "2025-01-17T01:23:21.766Z" },
+]
+
 [[package]]
 name = "fastapi"
 version = "0.136.3"
@@ -1165,6 +1216,10 @@ transformers = [
     { name = "torch" },
     { name = "transformers" },
 ]
+tts = [
+    { name = "kokoro-onnx" },
+    { name = "soundfile" },
+]
 
 [package.dev-dependencies]
 dev = [
@@ -1180,6 +1235,7 @@ requires-dist = [
     { name = "faster-whisper", marker = "extra == 'stt'", specifier = ">=1.0" },
     { name = "gradio", specifier = ">=5.0" },
     { name = "huggingface-hub", specifier = ">=1.17.0" },
+    { name = "kokoro-onnx", marker = "extra == 'tts'", specifier = ">=0.3" },
     { name = "llama-cpp-python", marker = "extra == 'llamacpp'", specifier = ">=0.3.0" },
     { name = "matplotlib", marker = "extra == 'dev'", specifier = ">=3.9" },
     { name = "modal", marker = "extra == 'modal'", specifier = ">=1.0" },
@@ -1193,6 +1249,7 @@ requires-dist = [
     { name = "ruff", marker = "extra == 'dev'", specifier = ">=0.6" },
     { name = "safetensors", marker = "extra == 'image'", specifier = ">=0.4" },
     { name = "sounddevice", marker = "extra == 'stt'", specifier = ">=0.5.5" },
+    { name = "soundfile", marker = "extra == 'tts'", specifier = ">=0.12" },
     { name = "spaces", marker = "extra == 'spaces'", specifier = ">=0.30" },
     { name = "torch", marker = "extra == 'image'", specifier = ">=2.3" },
     { name = "torch", marker = "extra == 'matting'", specifier = ">=2.3" },
@@ -1201,7 +1258,7 @@ requires-dist = [
     { name = "transformers", marker = "extra == 'matting'", specifier = ">=4.44" },
     { name = "transformers", marker = "extra == 'transformers'", specifier = ">=4.44" },
 ]
-provides-extras = ["llamacpp", "transformers", "image", "stt", "matting", "spaces", "modal", "dev"]
+provides-extras = ["llamacpp", "transformers", "image", "stt", "tts", "matting", "spaces", "modal", "dev"]
 
 [package.metadata.requires-dev]
 dev = [
@@ -1364,6 +1421,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/cb/b1/3846dd7f199d53cb17f49cba7e651e9ce294d8497c8c150530ed11865bb8/iniconfig-2.3.0-py3-none-any.whl", hash = "sha256:f631c04d2c48c52b84d0d0549c99ff3859c98df65b3101406327ecc7d53fbf12", size = 7484, upload-time = "2025-10-18T21:55:41.639Z" },
 ]
 
+[[package]]
+name = "isodate"
+version = "0.7.2"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/54/4d/e940025e2ce31a8ce1202635910747e5a87cc3a6a6bb2d00973375014749/isodate-0.7.2.tar.gz", hash = "sha256:4cd1aa0f43ca76f4a6c6c0292a85f40b35ec2e43e315b59f06e6d32171a953e6", size = 29705, upload-time = "2024-10-08T23:04:11.5Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/15/aa/0aca39a37d3c7eb941ba736ede56d689e7be91cab5d9ca846bde3999eba6/isodate-0.7.2-py3-none-any.whl", hash = "sha256:28009937d8031054830160fce6d409ed342816b543597cece116d966c6d99e15", size = 22320, upload-time = "2024-10-08T23:04:09.501Z" },
+]
+
 [[package]]
 name = "jinja2"
 version = "3.1.6"
@@ -1376,6 +1442,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/62/a1/3d680cbfd5f4b8f15abc1d571870c5fc3e594bb582bc3b64ea099db13e56/jinja2-3.1.6-py3-none-any.whl", hash = "sha256:85ece4451f492d0c13c5dd7c13a64681a86afae63a5f347908daf103ce6d2f67", size = 134899, upload-time = "2025-03-05T20:05:00.369Z" },
 ]
 
+[[package]]
+name = "joblib"
+version = "1.5.3"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/41/f2/d34e8b3a08a9cc79a50b2208a93dce981fe615b64d5a4d4abee421d898df/joblib-1.5.3.tar.gz", hash = "sha256:8561a3269e6801106863fd0d6d84bb737be9e7631e33aaed3fb9ce5953688da3", size = 331603, upload-time = "2025-12-15T08:41:46.427Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/7b/91/984aca2ec129e2757d1e4e3c81c3fcda9d0f85b74670a094cc443d9ee949/joblib-1.5.3-py3-none-any.whl", hash = "sha256:5fc3c5039fc5ca8c0276333a188bbd59d6b7ab37fe6632daa76bc7f9ec18e713", size = 309071, upload-time = "2025-12-15T08:41:44.973Z" },
+]
+
 [[package]]
 name = "jsonschema"
 version = "4.26.0"
@@ -1509,6 +1584,30 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/0a/dd/8050c947d435c8d4bc94e3252f4d8bb8a76cfb424f043a8680be637a57f1/kiwisolver-1.5.0-pp311-pypy311_pp73-win_amd64.whl", hash = "sha256:59cd8683f575d96df5bb48f6add94afc055012c29e28124fcae2b63661b9efb1", size = 73558, upload-time = "2026-03-09T13:15:52.112Z" },
 ]
 
+[[package]]
+name = "kokoro-onnx"
+version = "0.5.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "espeakng-loader" },
+    { name = "numpy" },
+    { name = "onnxruntime" },
+    { name = "phonemizer-fork" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/6b/18/277bd18aeeceaf5c46a51bb50c1cbdccdad8cab7fd1d58f0173bbeeec708/kokoro_onnx-0.5.0.tar.gz", hash = "sha256:5beb15f085e2828ed8d493f792c079af857103ab2dceaa1e112b1760587ac96a", size = 84570, upload-time = "2026-01-30T03:05:45.6Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0d/55/0bfcb4aa50033c89e5ac132af3d07fac0543824ce6eaefd4d1bfdcc3795b/kokoro_onnx-0.5.0-py3-none-any.whl", hash = "sha256:4e1c38a296db5dbc1f722f69f5e3c13f2e2877b3e5b145287f56ec057013e357", size = 17448, upload-time = "2026-01-30T03:05:46.931Z" },
+]
+
+[[package]]
+name = "language-tags"
+version = "1.3.1"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/39/d1/b37165bed9016c19e2cdecf4176fbd902a014fb8a87dd44415438e190969/language_tags-1.3.1.tar.gz", hash = "sha256:b15f05505dad3ad296a1782d5a6083fd141309186094d1ab08095348f4203f37", size = 222642, upload-time = "2026-05-08T11:46:24.418Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/0a/21/b48b9e8408b3faec9fb7cb2f68352c7724add35a980c948eaceb29ac41e4/language_tags-1.3.1-py3-none-any.whl", hash = "sha256:f7db7ef8879523019603e09644a86d95ba60595ce5e5ea82d46e8971b0f68f7b", size = 216654, upload-time = "2026-05-08T11:46:17.836Z" },
+]
+
 [[package]]
 name = "lazy-loader"
 version = "0.5"
@@ -2337,6 +2436,22 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/e8/b6/f54d676ed93cc2dd2234c3b172ea9c8c3d7d29361e66b1b23dec57a67465/peft-0.19.1-py3-none-any.whl", hash = "sha256:2113f72a81621b5913ef28f9022204c742df111890c5f49d812716a4a301e356", size = 680692, upload-time = "2026-04-16T15:46:42.886Z" },
 ]
 
+[[package]]
+name = "phonemizer-fork"
+version = "3.3.2"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "attrs" },
+    { name = "dlinfo" },
+    { name = "joblib" },
+    { name = "segments" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/42/fa/9294d2f11890ca49d0bdac7a4da60cbe5686629bfd4987cae0ad75e051cc/phonemizer_fork-3.3.2.tar.gz", hash = "sha256:10e16e827d0443b087062e21b55e805c00989cf1343b2e81e734cae5f6c0cf69", size = 300989, upload-time = "2025-01-30T13:02:31.201Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/64/f1/0dcce21b0ae16a82df4b6583f8f3ad8e55b35f7e98b6bf536a4dd225fa08/phonemizer_fork-3.3.2-py3-none-any.whl", hash = "sha256:97305c76f4183b3825dae8f4c032265fe78c9946ce58c47d4b62161349264b74", size = 82700, upload-time = "2025-01-30T13:02:28.667Z" },
+]
+
 [[package]]
 name = "pillow"
 version = "12.2.0"
@@ -2888,6 +3003,18 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/f1/12/de94a39c2ef588c7e6455cfbe7343d3b2dc9d6b6b2f40c4c6565744c873d/pyyaml-6.0.3-cp314-cp314t-win_arm64.whl", hash = "sha256:ebc55a14a21cb14062aa4162f906cd962b28e2e9ea38f9b4391244cd8de4ae0b", size = 149341, upload-time = "2025-09-25T21:32:56.828Z" },
 ]
 
+[[package]]
+name = "rdflib"
+version = "7.6.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "pyparsing" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/98/f5/18bb77b7af9526add0c727a3b2048959847dc5fb030913e2918bf384fec3/rdflib-7.6.0.tar.gz", hash = "sha256:6c831288d5e4a5a7ece85d0ccde9877d512a3d0f02d7c06455d00d6d0ea379df", size = 4943826, upload-time = "2026-02-13T07:15:55.938Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/10/c2/6604a71269e0c1bd75656d5a001432d16f2cc5b8c057140ec797155c295e/rdflib-7.6.0-py3-none-any.whl", hash = "sha256:30c0a3ebf4c0e09215f066be7246794b6492e054e782d7ac2a34c9f70a15e0dd", size = 615416, upload-time = "2026-02-13T07:15:46.487Z" },
+]
+
 [[package]]
 name = "referencing"
 version = "0.37.0"
@@ -3040,6 +3167,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/a0/f4/c67b0b3f1b9245e8d266f0f112c500d50e5b4e83cb6f3b71b6528104182a/requests-2.34.2-py3-none-any.whl", hash = "sha256:2a0d60c172f83ac6ab31e4554906c0f3b3588d37b5cb939b1c061f4907e278e0", size = 73075, upload-time = "2026-05-14T19:25:26.443Z" },
 ]
 
+[[package]]
+name = "rfc3986"
+version = "1.5.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/79/30/5b1b6c28c105629cc12b33bdcbb0b11b5bb1880c6cfbd955f9e792921aa8/rfc3986-1.5.0.tar.gz", hash = "sha256:270aaf10d87d0d4e095063c65bf3ddbc6ee3d0b226328ce21e036f946e421835", size = 49378, upload-time = "2021-05-07T23:29:27.183Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/c4/e5/63ca2c4edf4e00657584608bee1001302bbf8c5f569340b78304f2f446cb/rfc3986-1.5.0-py2.py3-none-any.whl", hash = "sha256:a86d6e1f5b1dc238b218b012df0aa79409667bb209e58da56d0b94704e712a97", size = 31976, upload-time = "2021-05-07T23:29:25.611Z" },
+]
+
 [[package]]
 name = "rich"
 version = "15.0.0"
@@ -3389,6 +3525,19 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/07/39/338d9219c4e87f3e708f18857ecd24d22a0c3094752393319553096b98af/scipy-1.17.1-cp314-cp314t-win_arm64.whl", hash = "sha256:200e1050faffacc162be6a486a984a0497866ec54149a01270adc8a59b7c7d21", size = 25489165, upload-time = "2026-02-23T00:22:29.563Z" },
 ]
 
+[[package]]
+name = "segments"
+version = "2.4.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "csvw" },
+    { name = "regex" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/8b/57/85cac3a8e32370e88fa5fa92812edb6025db7fcbed51452bd56ee1524957/segments-2.4.0.tar.gz", hash = "sha256:bba71f5520ddd54c8aa2f4d765a60618c6862162d6e7356a4a097f2223166f5b", size = 18662, upload-time = "2026-03-07T10:01:28.925Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/be/60/eef9acce946177f92c9aabf432224d87ab908bafafac516a36ab924199f3/segments-2.4.0-py2.py3-none-any.whl", hash = "sha256:4021dc67f201cc03c864c74c618bdb163b1af629da3040babbaa37d8813f3db0", size = 16321, upload-time = "2026-03-07T10:01:27.885Z" },
+]
+
 [[package]]
 name = "semantic-version"
 version = "2.10.0"
@@ -3441,6 +3590,27 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/4e/39/a61d4b83a7746b70d23d9173be688c0c6bfc7173772344b7442c2c155497/sounddevice-0.5.5-py3-none-win_arm64.whl", hash = "sha256:3861901ddd8230d2e0e8ae62ac320cdd4c688d81df89da036dcb812f757bb3e6", size = 317115, upload-time = "2026-01-23T18:36:42.235Z" },
 ]
 
+[[package]]
+name = "soundfile"
+version = "0.14.0"
+source = { registry = "https://pypi.org/simple" }
+dependencies = [
+    { name = "cffi" },
+    { name = "numpy" },
+    { name = "typing-extensions" },
+]
+sdist = { url = "https://files.pythonhosted.org/packages/d2/db/949331952a6fb1c5b12e9de80fd08747966c2039d1a61db4764fbd3981c2/soundfile-0.14.0.tar.gz", hash = "sha256:ba1c1a2d618bca5c406647c83b89f07cc8810fa506a50622a6993ba130c1de11", size = 47842, upload-time = "2026-06-06T08:58:47.869Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/b1/d1/5e338af9ca6ed0786cd5bb03f6d60de1c325728c1189014f3b59aae7403c/soundfile-0.14.0-py2.py3-none-any.whl", hash = "sha256:8ba81ae3a89fd5ab3bef8a8eb481fbbe794e806309675a89b4df48b8d31908a8", size = 26799, upload-time = "2026-06-06T08:58:33.269Z" },
+    { url = "https://files.pythonhosted.org/packages/7e/72/c6b21e58d3113596e7e8de0a08d6f1d95173492cfbca0a4db14148cbba2a/soundfile-0.14.0-py2.py3-none-macosx_10_9_x86_64.whl", hash = "sha256:19be05428da76ed61a4cad29b8e4bcf43a3e5c100089d2ec81dc961eed1b0dd4", size = 1144568, upload-time = "2026-06-06T08:58:35.231Z" },
+    { url = "https://files.pythonhosted.org/packages/63/7a/dfdd6f8c748988427119f75eb860a3cedd858d1aea1fe28f39ad8559ef22/soundfile-0.14.0-py2.py3-none-macosx_11_0_arm64.whl", hash = "sha256:d828d35a059626da52f1415b5faee610aeab393319cb3fc4a9aef47b619fc14c", size = 1103726, upload-time = "2026-06-06T08:58:37.948Z" },
+    { url = "https://files.pythonhosted.org/packages/4a/f8/fc39fad6f879633461d27394cd1ddaf1f769ffa0597dca35872f51b16461/soundfile-0.14.0-py2.py3-none-manylinux_2_28_aarch64.whl", hash = "sha256:e85724a90bc99a6e8062c0b4ddf725f53b2a3b70afd4da875e9d2cfc4e92f377", size = 1238050, upload-time = "2026-06-06T08:58:39.932Z" },
+    { url = "https://files.pythonhosted.org/packages/7b/a2/70fd4432b924684c372df8b0a45708c36c057ef3596c9eb53e0a806b980b/soundfile-0.14.0-py2.py3-none-manylinux_2_28_x86_64.whl", hash = "sha256:1e38bac1853412871318e82a1ba69a8be677619b56025bbfcccdb41b6cafe82d", size = 1315963, upload-time = "2026-06-06T08:58:41.716Z" },
+    { url = "https://files.pythonhosted.org/packages/d9/34/c9e80783d83eab739a9531fdee03675d53e0bf1b2ccb4bb3af5844675046/soundfile-0.14.0-py2.py3-none-win32.whl", hash = "sha256:0a6ae43c50c71b4e020cc55382925cb89451c1ed1a0c3d0f5d802da269226849", size = 902199, upload-time = "2026-06-06T08:58:43.289Z" },
+    { url = "https://files.pythonhosted.org/packages/ed/97/b39c18ac1df45e755ca22b8b00e872929da5d107998a207a5e4ac831bfda/soundfile-0.14.0-py2.py3-none-win_amd64.whl", hash = "sha256:299491d3499460fb1b74bb4bd78b57ffc2d243a5fafa7b6ec1b264875c78453e", size = 1021480, upload-time = "2026-06-06T08:58:45.016Z" },
+    { url = "https://files.pythonhosted.org/packages/f4/83/55c65e61cf457805ce2ec157c1c6ae17715d0851aa2374422de0538838ca/soundfile-0.14.0-py2.py3-none-win_arm64.whl", hash = "sha256:e090704718e124e7c844695236f1fce8d18a5e761eaf7c82dfcd124620805f98", size = 888858, upload-time = "2026-06-06T08:58:46.593Z" },
+]
+
 [[package]]
 name = "spaces"
 version = "0.50.2"
@@ -3495,6 +3665,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/57/ea/531a6ea751cbd989da386144810b1b8f529b0aae8c1a9beda8b40966c9c2/synchronicity-0.12.3-py3-none-any.whl", hash = "sha256:e476818cd14102136f41622c619de548f0000c024485fc18521c8fe908ea7574", size = 40982, upload-time = "2026-05-28T12:33:49.125Z" },
 ]
 
+[[package]]
+name = "termcolor"
+version = "3.3.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/46/79/cf31d7a93a8fdc6aa0fbb665be84426a8c5a557d9240b6239e9e11e35fc5/termcolor-3.3.0.tar.gz", hash = "sha256:348871ca648ec6a9a983a13ab626c0acce02f515b9e1983332b17af7979521c5", size = 14434, upload-time = "2025-12-29T12:55:21.882Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/33/d1/8bb87d21e9aeb323cc03034f5eaf2c8f69841e40e4853c2627edf8111ed3/termcolor-3.3.0-py3-none-any.whl", hash = "sha256:cf642efadaf0a8ebbbf4bc7a31cec2f9b5f21a9f726f4ccbb08192c9c26f43a5", size = 7734, upload-time = "2025-12-29T12:55:20.718Z" },
+]
+
 [[package]]
 name = "tifffile"
 version = "2026.3.3"
@@ -3741,6 +3920,15 @@ wheels = [
     { url = "https://files.pythonhosted.org/packages/ce/e4/dccd7f47c4b64213ac01ef921a1337ee6e30e8c6466046018326977efd95/tzdata-2026.2-py2.py3-none-any.whl", hash = "sha256:bbe9af844f658da81a5f95019480da3a89415801f6cc966806612cc7169bffe7", size = 349321, upload-time = "2026-04-24T15:22:05.876Z" },
 ]
 
+[[package]]
+name = "uritemplate"
+version = "4.2.0"
+source = { registry = "https://pypi.org/simple" }
+sdist = { url = "https://files.pythonhosted.org/packages/98/60/f174043244c5306c9988380d2cb10009f91563fc4b31293d27e17201af56/uritemplate-4.2.0.tar.gz", hash = "sha256:480c2ed180878955863323eea31b0ede668795de182617fef9c6ca09e6ec9d0e", size = 33267, upload-time = "2025-06-02T15:12:06.318Z" }
+wheels = [
+    { url = "https://files.pythonhosted.org/packages/a9/99/3ae339466c9183ea5b8ae87b34c0b897eda475d2aec2307cae60e5cd4f29/uritemplate-4.2.0-py3-none-any.whl", hash = "sha256:962201ba1c4edcab02e60f9a0d3821e82dfc5d2d6662a21abd533879bdb8a686", size = 11488, upload-time = "2025-06-02T15:12:03.405Z" },
+]
+
 [[package]]
 name = "urllib3"
 version = "2.7.0"

visualnovel/config.py

@@ -44,15 +44,20 @@ USE_MOCK = os.getenv("VN_MOCK", "1") == "1"
 # VN_REAL_STT=1 -> use real Whisper even in mock mode (test STT without full model stack).
 REAL_STT = os.getenv("VN_REAL_STT", "0") == "1"
 
-# LLM runtime backend when not mocking.  llamacpp = the llama.cpp bonus path.
-LLM_BACKEND = os.getenv("VN_LLM_BACKEND", "llamacpp")  # llamacpp | transformers | modal | mock
+# Detect HuggingFace Space runtime — SPACE_ID is always injected by the HF platform.
+_ON_SPACE = bool(os.getenv("SPACE_ID"))
+
+# LLM runtime backend when not mocking.
+# Auto-selects transformers on HF Spaces (ZeroGPU); llamacpp everywhere else.
+# Override with VN_LLM_BACKEND=modal for the Modal cloud path.
+LLM_BACKEND = os.getenv("VN_LLM_BACKEND", "transformers" if _ON_SPACE else "llamacpp")
 
 # Image generation backend when not mocking.
-# Defaults to "modal" when LLM_BACKEND is already modal, otherwise "local" (SDXL-Turbo on-device).
+# "lightning" = SDXL-Lightning (4-step, GPU-local)  |  "modal" = Modal cloud  |  "local" = SDXL-Turbo
 IMAGE_BACKEND = os.getenv(
     "VN_IMAGE_BACKEND",
-    "modal" if LLM_BACKEND == "modal" else "local",
-)  # modal | local
+    "modal" if LLM_BACKEND == "modal" else "lightning",
+)
 
 # Set to a .jsonl path to record a full per-turn trace (Open-Trace bonus).
 TRACE_PATH = os.getenv("VN_TRACE")  # None disables tracing
@@ -76,6 +81,8 @@ WHISPER_SIZE = os.getenv(
     "VN_WHISPER_SIZE", "large-v3-turbo"
 )  # tiny|base|small|medium|large-v3|large-v3-turbo
 
+TTS_BACKEND = os.getenv("VN_TTS_BACKEND", "mock")  # mock | kokoro
+
 # --------------------------------------------------------------------------- #
 #  Generation / budget params
 # --------------------------------------------------------------------------- #

visualnovel/engine.py

@@ -14,6 +14,7 @@ sessions on a busy Space, key engines by session id instead.
 
 from __future__ import annotations
 
+import base64
 from pathlib import Path
 
 from . import config, memory, orchestrator, state
@@ -23,6 +24,7 @@ from .painter import get_painter
 from .schemas import DirectorOutput, GameState, SetupForm, SpritePresence, Turn, ViewState
 from .stt import get_stt
 from .trace import Tracer
+from .tts import get_tts
 
 
 class Engine:
@@ -30,6 +32,7 @@ class Engine:
         self.llm = get_llm()
         self.painter = get_painter()
         self.stt = get_stt()
+        self.tts = get_tts()
         self.tracer = Tracer(config.TRACE_PATH)
         self.state: GameState | None = None
 
@@ -165,7 +168,18 @@ class Engine:
                 )
             )
 
-        speaker_name = s.characters[out.speaker].name if out.speaker in s.characters else "The wood"
+        speaker_ch = s.characters.get(out.speaker)
+        speaker_name = speaker_ch.name if speaker_ch else "The wood"
+
+        audio_b64: str | None = None
+        if speaker_ch and speaker_ch.tts_voice_description:
+            with collector.measure("tts", turn=turn):
+                wav = self.tts.synthesize(
+                    out.dialogue, speaker_ch.tts_voice_description, speaker_ch.sprite_seed
+                )
+            if wav:
+                audio_b64 = "data:audio/wav;base64," + base64.b64encode(wav).decode()
+
         return ViewState(
             speaker=speaker_name,
             dialogue=out.dialogue,
@@ -180,6 +194,7 @@ class Engine:
             turn_index=s.turn_index,
             notifications=notifications or [],
             intro_text=intro_text,
+            audio_b64=audio_b64,
         )
 
 

visualnovel/orchestrator.py

@@ -93,6 +93,7 @@ def init_world(llm: LLMBackend, setup: SetupForm) -> tuple[GameState, DirectorOu
                 traits=first.traits,
                 goals=first.goals,
                 sprite_seed=_seed_for(seed, first.id),
+                tts_voice_description=first.voice,  # frozen at creation
             )
         },
         flags=flags,

visualnovel/schemas.py

@@ -40,6 +40,7 @@ class Character(BaseModel):
     # Discovery tracking (not sent to the LLM — managed by state.apply_directives)
     discovered_traits: list[int] = Field(default_factory=list)  # indices of unlocked traits
     goal_unlocked: bool = False
+    tts_voice_description: str = ""  # frozen at first introduction; never updated by directives
 
 
 class Scene(BaseModel):
@@ -166,3 +167,4 @@ class ViewState(BaseModel):
     turn_index: int = 0
     notifications: list[str] = Field(default_factory=list)  # popup messages this turn
     intro_text: str = ""  # shown once in the book overlay at game start; empty on subsequent turns
+    audio_b64: str | None = None  # data:audio/wav;base64,... — absent when TTS is mocked

visualnovel/state.py

@@ -63,6 +63,7 @@ def apply_directives(state: GameState, out: DirectorOutput) -> list[str]:
                 traits=nc.traits,
                 goals=nc.goals,
                 sprite_seed=_seed_for(state.seed, cid),
+                tts_voice_description=nc.voice,  # frozen at creation; never updated
             )
         if cid not in state.scene.present:
             state.scene.present.append(cid)

visualnovel/tts.py

@@ -0,0 +1,151 @@
+"""The Voice — text to speech.
+
+Synthesises NPC dialogue using Kokoro-82M via ONNX (~82M params, no transformers dependency).
+
+Design notes:
+- Voice selection reads Character.tts_voice_description (frozen at creation) and scores every
+  Kokoro preset by keyword overlap — the best-scoring voice wins.  sprite_seed breaks ties so
+  two characters with the same description still get different voices.
+- `tts_voice_description` is frozen at character creation (see state.apply_directives and
+  orchestrator.init_world), ensuring voice identity is stable for the whole session.
+- Results are cached in memory by SHA-256(text + voice_id + seed) so repeated lines are free.
+
+Model files (~337 MB total) live in models/kokoro/ and are downloaded once:
+  curl -L -o models/kokoro/kokoro-v1.0.onnx \\
+    https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/kokoro-v1.0.onnx
+  curl -L -o models/kokoro/voices-v1.0.bin \\
+    https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/voices-v1.0.bin
+"""
+
+from __future__ import annotations
+
+import hashlib
+import io
+from typing import Protocol
+
+from . import config
+
+# ---------------------------------------------------------------------------
+# Voice profiles — (voice_id, gender, keyword_set)
+# gender: "f" | "m"
+# keyword_set: words in a voice descriptor that suggest this voice
+# ---------------------------------------------------------------------------
+_VOICE_PROFILES: list[tuple[str, str, frozenset[str]]] = [
+    # ── American Female ────────────────────────────────────────────────────
+    ("af_heart",   "f", frozenset({"warm", "bright", "breathless", "tender", "sweet", "laughs", "light"})),
+    ("af_bella",   "f", frozenset({"soft", "gentle", "quiet", "shy", "delicate", "whisper", "whispered"})),
+    ("af_sarah",   "f", frozenset({"clear", "professional", "calm", "steady", "cool", "composed"})),
+    ("af_sky",     "f", frozenset({"airy", "bubbly", "cheerful", "energetic", "lively", "peppy"})),
+    ("af_jessica", "f", frozenset({"bold", "confident", "sharp", "spirited", "assertive", "direct"})),
+    ("af_nicole",  "f", frozenset({"rich", "smooth", "soothing", "mature", "warm", "velvet"})),
+    ("af_nova",    "f", frozenset({"crisp", "precise", "intelligent", "focused", "analytical"})),
+    ("af_river",   "f", frozenset({"flowing", "peaceful", "serene", "dreamy", "drifts", "trails"})),
+    ("af_kore",    "f", frozenset({"strong", "powerful", "serious", "stern", "commanding"})),
+    ("af_aoede",   "f", frozenset({"melodic", "musical", "lyrical", "poetic", "ethereal", "sings"})),
+    ("af_alloy",   "f", frozenset({"neutral", "balanced", "modern", "even"})),
+    # ── British Female ─────────────────────────────────────────────────────
+    ("bf_alice",    "f", frozenset({"british", "english", "uk", "posh", "accent", "clipped"})),
+    ("bf_emma",     "f", frozenset({"british", "english", "uk", "homely", "friendly", "warm"})),
+    ("bf_isabella", "f", frozenset({"british", "english", "uk", "refined", "elegant", "sophisticated"})),
+    ("bf_lily",     "f", frozenset({"british", "english", "uk", "gentle", "soft", "sweet"})),
+    # ── American Male ─────────────────────────────────────────────────────
+    ("am_adam",    "m", frozenset({"confident", "deep", "steady", "bass", "strong", "rich"})),
+    ("am_echo",    "m", frozenset({"resonant", "full", "commanding", "measured"})),
+    ("am_eric",    "m", frozenset({"friendly", "casual", "warm", "approachable", "cheerful"})),
+    ("am_fenrir",  "m", frozenset({"gruff", "rough", "intense", "dark", "rugged", "growl", "harsh"})),
+    ("am_liam",    "m", frozenset({"young", "energetic", "bright", "enthusiastic", "lively"})),
+    ("am_michael", "m", frozenset({"mature", "calm", "authoritative", "professorial", "quiet"})),
+    ("am_onyx",    "m", frozenset({"smooth", "velvet", "baritone", "low", "silky", "suave"})),
+    ("am_puck",    "m", frozenset({"playful", "mischievous", "quick", "witty", "impish", "teasing"})),
+    # ── British Male ──────────────────────────────────────────────────────
+    ("bm_daniel",  "m", frozenset({"british", "english", "uk", "smooth", "refined", "crisp"})),
+    ("bm_fable",   "m", frozenset({"british", "english", "uk", "storyteller", "dramatic", "expressive"})),
+    ("bm_george",  "m", frozenset({"british", "english", "uk", "distinguished", "formal", "noble"})),
+    ("bm_lewis",   "m", frozenset({"british", "english", "uk", "young", "casual", "easy"})),
+]
+
+_MALE_GENDER_HINTS   = {"man", "male", "his", "him", "he", "boy", "lad", "gentleman"}
+_FEMALE_GENDER_HINTS = {"woman", "female", "her", "she", "girl", "lady"}
+
+_MODEL_PATH  = config.MODELS_DIR / "kokoro" / "kokoro-v1.0.onnx"
+_VOICES_PATH = config.MODELS_DIR / "kokoro" / "voices-v1.0.bin"
+
+
+def _pick_voice(voice_description: str, sprite_seed: int) -> str:
+    """Score every Kokoro voice against the description; use sprite_seed to break ties."""
+    words = set(voice_description.lower().split())
+
+    # Detect gender from description
+    is_male   = bool(words & _MALE_GENDER_HINTS)
+    is_female = bool(words & _FEMALE_GENDER_HINTS)
+    # If ambiguous, default to female (most anime VN characters)
+    target_gender = "m" if is_male and not is_female else "f"
+
+    candidates = [(vid, kws) for vid, g, kws in _VOICE_PROFILES if g == target_gender]
+
+    # Score: number of keyword matches, then deterministic tie-break by seed
+    scored = [(len(words & kws), vid) for vid, kws in candidates]
+    scored.sort(key=lambda x: x[0], reverse=True)
+
+    # Among voices tied at the top score, pick by sprite_seed
+    top_score = scored[0][0]
+    top_voices = [vid for score, vid in scored if score == top_score]
+    return top_voices[sprite_seed % len(top_voices)]
+
+
+class TTSBackend(Protocol):
+    def synthesize(self, text: str, voice_description: str, seed: int) -> bytes | None: ...
+
+
+class MockTTS:
+    def synthesize(self, text: str, voice_description: str, seed: int) -> bytes | None:
+        return None
+
+
+class KokoroTTS:
+    def __init__(self) -> None:
+        from kokoro_onnx import Kokoro  # noqa: PLC0415
+
+        if not _MODEL_PATH.exists() or not _VOICES_PATH.exists():
+            raise FileNotFoundError(
+                f"Kokoro model files missing from {_MODEL_PATH.parent}.\n"
+                "Download them with:\n"
+                "  curl -L -o models/kokoro/kokoro-v1.0.onnx \\\n"
+                "    https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/kokoro-v1.0.onnx\n"
+                "  curl -L -o models/kokoro/voices-v1.0.bin \\\n"
+                "    https://github.com/thewh1teagle/kokoro-onnx/releases/download/model-files-v1.0/voices-v1.0.bin"
+            )
+
+        self._kokoro = Kokoro(str(_MODEL_PATH), str(_VOICES_PATH))
+        # Filter profiles to voices actually present in the loaded model
+        available = set(self._kokoro.get_voices())
+        self._available_voices = available
+        self._cache: dict[str, bytes] = {}
+
+    def synthesize(self, text: str, voice_description: str, seed: int) -> bytes | None:
+        import soundfile as sf  # noqa: PLC0415
+
+        voice_id = _pick_voice(voice_description, seed)
+        # Fallback if the chosen voice isn't in this model build
+        if voice_id not in self._available_voices:
+            fallback = [vid for vid, g, _ in _VOICE_PROFILES if g == "f" and vid in self._available_voices]
+            voice_id = fallback[seed % len(fallback)] if fallback else next(iter(self._available_voices))
+
+        cache_key = hashlib.sha256(f"{text}\x00{voice_id}\x00{seed}".encode()).hexdigest()
+        if cache_key in self._cache:
+            return self._cache[cache_key]
+
+        samples, sample_rate = self._kokoro.create(text, voice=voice_id, speed=1.0, lang="en-us")
+
+        buf = io.BytesIO()
+        sf.write(buf, samples, sample_rate, format="WAV")
+        wav_bytes = buf.getvalue()
+
+        self._cache[cache_key] = wav_bytes
+        return wav_bytes
+
+
+def get_tts() -> TTSBackend:
+    if config.TTS_BACKEND == "kokoro":
+        return KokoroTTS()
+    return MockTTS()