Update app.py

app.py

@@ -71,7 +71,6 @@ XTTS_MODEL.load_checkpoint(
 
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
-# налады
 torch.set_num_threads(1)
 if device.startswith("cuda"):
     torch.backends.cuda.matmul.allow_tf32 = True
@@ -87,7 +86,7 @@ tokenizer = VoiceBpeTokenizer(vocab_file=vocab_file)
 XTTS_MODEL.tokenizer = tokenizer
 
 # =========================================================
-# 3) Канстанты стриму (меншыя затрымкі)
+# 3) Канстанты стриму
 # =========================================================
 MIN_BUFFER_S = 0.020
 RUNTIME_FIRST_CHUNK_S = 0.010
@@ -97,7 +96,7 @@ ENABLE_TEXT_SPLITTING = True
 FIRST_SEGMENT_LIMIT = 120
 
 # =========================================================
-# 4) Утыліты аўдыя (стрим у INT16 base64)
+# 4) Аўдыя-ўтыліты
 # =========================================================
 def _seconds_to_samples(sec: float, sr: int) -> int:
     return max(1, int(sec * sr))
@@ -165,7 +164,7 @@ def _pcm_f32_to_int16_b64(x: np.ndarray) -> str:
     if x.dtype != np.float32:
         x = x.astype(np.float32, copy=False)
     y = np.clip(x, -1.0, 0.9999695)
-    i16 = (y * 32767.0).astype("<i2", copy=False)  # little-endian int16
+    i16 = (y * 32767.0).astype("<i2", copy=False)
     return base64.b64encode(i16.tobytes()).decode("ascii")
 
 # =========================================================
@@ -394,7 +393,7 @@ except Exception as e:
     print(f"[warn] warm-up failed: {e}")
 
 # =========================================================
-# 7) Падзел тэксту (хуткі + fallback)
+# 7) Падзел тэксту
 # =========================================================
 _SENT_END = re.compile(r'([\.!\?…]+[»")\]]*\s+)')
 _WS = re.compile(r"\s+")
@@ -469,16 +468,13 @@ def _split_text_smart(text_in: str, lang_short: str, chunk_limit: int) -> List[s
     return parts + (rest or [text_for_rest])
 
 # =========================================================
-# 8) TTS стрим + лагі
-#    — замянілі механіку: Python выдае JSON-пакеты ў адзін схаваны Textbox,
-#      фронт (JS у gr.HTML) опускае іх праз polling у AudioWorklet
+# 8) TTS стрим (выдае JSON-пакеты ў Textbox)
 # =========================================================
 @spaces.GPU(duration=60)
 def text_to_speech(belarusian_story, speaker_audio_file=None):
     t0 = time.perf_counter()
 
     if not belarusian_story or str(belarusian_story).strip() == "":
-        # пусты пакет, каб фронт не зламаўся
         yield (json.dumps({"seq": 0, "b64": "", "log": None, "stop": False}), None, None)
         raise gr.Error("Увядзі хоць нейкі тэкст 🙂")
 
@@ -491,7 +487,7 @@ def text_to_speech(belarusian_story, speaker_audio_file=None):
     lang_short = "be"
     chunk_limit = getattr(XTTS_MODEL.tokenizer, "char_limits", {}).get(lang_short, 250)
 
-    # Latents (кэш CPU/GPU)
+    # Latents
     t_lat0 = time.perf_counter()
     to_dev = device if device.startswith("cuda") else None
     gpt_cond_latent, speaker_embedding = _latents_for(speaker_audio_file, to_device=to_dev)
@@ -514,7 +510,6 @@ def text_to_speech(belarusian_story, speaker_audio_file=None):
     }
 
     seq = 0
-    # пачатковы пакет — паведамляем фронту, што ўсё гатова
     yield (json.dumps({"seq": seq, "b64": "", "log": server_metrics, "stop": False}), None, None)
 
     full_audio_chunks: List[np.ndarray] = []
@@ -568,18 +563,16 @@ def text_to_speech(belarusian_story, speaker_audio_file=None):
             final_file_path = tmp.name
             final_audio_path = tmp.name
         except Exception as e:
-            # паведамляем пра памылку праз лаг
             server_metrics["_file_error"] = str(e)
         finally:
             t_w1 = time.perf_counter()
             server_metrics["file_write_s"] = (t_w1 - t_w0)
 
-    # стоп-пакет
     seq += 1
     yield (json.dumps({"seq": seq, "b64": "__STOP__", "log": server_metrics, "stop": True}), final_file_path, final_audio_path)
 
 # =========================================================
-# 9) UI + AudioWorklet з polling (без SharedArrayBuffer, без gr.Audio streaming)
+# 9) UI + AudioWorklet з polling (без gr.Audio streaming)
 # =========================================================
 examples = [
     [
@@ -606,12 +599,19 @@ with gr.Blocks() as demo:
         label="Лагі плэера",
     )
 
-    # Схаваныя каналы: stream JSON + выніковы файл + аўдыя
-    stream_pipe = gr.Textbox(value="", visible=False, label="stream_pipe", elem_id="stream-pipe")
+    # ВАЖНА: робім stream_pipe бачным у DOM (visible=True), але хаваем праз CSS
+    stream_pipe = gr.Textbox(value="", visible=True, label="stream_pipe", elem_id="stream-pipe")
     final_file = gr.File(label="Згенераваны WAV (спампаваць)")
     final_audio = gr.Audio(label="Фінальнае аўдыя", type="filepath", interactive=False, elem_id="final-audio")
     play_final_btn = gr.Button("▶️ Play Final")
 
+    # CSS — схаваць stream_pipe і актыўнасці
+    gr.HTML("""
+<style>
+#stream-pipe { position:absolute; left:-99999px; width:1px; height:1px; opacity:0; pointer-events:none; }
+</style>
+""")
+
     # --------- Frontend JS (пастаянны polling + AudioWorklet) ----------
     FRONT_HTML = f"""
 <script>
@@ -623,7 +623,6 @@ with gr.Blocks() as demo:
   function toSec(ms) {{ return (ms/1000); }}
   function fmtS(x) {{ return (x==null) ? 'n/a' : (x.toFixed ? x.toFixed(3) : x) + ' s'; }}
 
-  // Лагі
   function updateLog() {{
     const el = document.getElementById('wa-log');
     if (!el || !window.__wa || !window.__wa.meta) return;
@@ -664,13 +663,22 @@ with gr.Blocks() as demo:
   }}
 
   async function ensureWorklet(ctx) {{
-    // AudioWorklet код у выглядзе радка
     const code = `
 class PushPlayerProcessor extends AudioWorkletProcessor {{
   constructor() {{
     super();
     this.queue = [];
     this.readIndex = 0;
+    this.port.onmessage = (e) => {{
+      const d = e.data || {{}};
+      if (d.type === 'push' && d.buffer) {{
+        const f32 = new Float32Array(d.buffer);
+        this.queue.push(f32);
+      }} else if (d.type === 'reset') {{
+        this.queue.length = 0;
+        this.readIndex = 0;
+      }}
+    }};
   }}
   process(inputs, outputs) {{
     const out = outputs[0][0];
@@ -694,7 +702,6 @@ registerProcessor('push-player', PushPlayerProcessor);
     await ctx.audioWorklet.addModule(url);
   }}
 
-  // Ініцыялізацыя плэера
   async function ensurePlayer() {{
     if (window.__wa) return window.__wa;
     if (!AC) return null;
@@ -705,7 +712,7 @@ registerProcessor('push-player', PushPlayerProcessor);
     node.connect(ctx.destination);
 
     let playing = true;
-    let seq_seen = -1;
+
     const meta = {{
       t_click_ms: null,
       t_first_push_ms: null,
@@ -716,20 +723,19 @@ registerProcessor('push-player', PushPlayerProcessor);
     const api = {{
       ctx, node,
       get playing() {{ return playing; }},
-      start: async () => {{ try {{ await ctx.resume(); }} catch(e) {{}} playing = true; updateLog(); }},
-      stop: () => {{ playing = false; updateLog(); }},
+      start: async () => {{ try {{ await ctx.resume(); }} catch(e) {{}} playing = True; updateLog(); }},
+      stop: () => {{ try {{ ctx.suspend(); }} catch(e){{}} playing = false; updateLog(); }},
       reset: () => {{
-        seq_seen = -1;
+        try {{ node.port.postMessage({{ type: 'reset' }}); }} catch(e) {{}}
         meta.t_first_push_ms = null;
         meta.t_first_audio_ms = null;
         updateLog();
       }},
-      push: (f32) => {{
+      push: (f32) {{
         try {{ node.port.postMessage({{ type: 'push', buffer: f32.buffer }}, [f32.buffer]); }} catch (e) {{}}
         if (!meta.t_first_push_ms) {{
           meta.t_first_push_ms = performance.now();
-          // адразу адзначым пачатак гуку
-          if (!meta.t_first_audio_ms && playing) meta.t_first_audio_ms = meta.t_first_push_ms + 10;
+          if (!meta.t_first_audio_ms) meta.t_first_audio_ms = meta.t_first_push_ms + 10;
           updateLog();
         }}
         if (!playing) api.start();
@@ -741,26 +747,35 @@ registerProcessor('push-player', PushPlayerProcessor);
     return api;
   }}
 
-  // Polling чытанне з скрытага Textbox
-  async function pollLoop() {{
-    const api = await ensurePlayer();
-    if (!api) return;
-    const box = document.querySelector('#stream-pipe textarea, #stream-pipe input') || document.getElementById('stream-pipe');
-    if (!box) return;
+  function getPipeEl() {{
+    // Textbox у Gradio мае textarea унутры div#stream-pipe
+    const root = document.getElementById('stream-pipe');
+    if (!root) return null;
+    const ta = root.querySelector('textarea');
+    if (ta) return ta;
+    const inp = root.querySelector('input');
+    if (inp) return inp;
+    return root;
+  }}
 
-    function parsePacket(txt) {{
-      try {{ return JSON.parse(txt); }} catch(e) {{ return null; }}
-    }}
+  function startPolling() {{
+    const pipe = getPipeEl();
+    if (!pipe) return;
+    const apiPromise = ensurePlayer();
+    let lastSeq = -1;
 
-    // асноўны цыкл
-    setInterval(() => {{
-      const txt = (box.value !== undefined) ? box.value : (box.innerText || box.textContent || '');
-      const pkt = parsePacket(txt);
-      if (!pkt || typeof pkt.seq !== 'number') return;
-      if (!api.meta.t_click_ms) api.meta.t_click_ms = performance.now();
+    setInterval(async () => {{
+      const api = await apiPromise;
+      if (!api) return;
 
-      if (pkt.seq <= (api._seq_seen || -1)) return;
-      api._seq_seen = pkt.seq;
+      const txt = (pipe.value !== undefined) ? pipe.value : (pipe.innerText || pipe.textContent || '');
+      if (!txt) return;
+
+      let pkt = null;
+      try {{ pkt = JSON.parse(txt); }} catch(e) {{ return; }}
+      if (!pkt || typeof pkt.seq !== 'number') return;
+      if (pkt.seq <= lastSeq) return;
+      lastSeq = pkt.seq;
 
       if (pkt.log) {{
         api.meta.server = pkt.log;
@@ -774,7 +789,6 @@ registerProcessor('push-player', PushPlayerProcessor);
       }}
 
       if (typeof pkt.b64 === 'string' && pkt.b64.length > 0) {{
-        // распакаваць INT16LE->F32
         const bin = atob(pkt.b64);
         const len = bin.length;
         const buf = new ArrayBuffer(len);
@@ -791,7 +805,6 @@ registerProcessor('push-player', PushPlayerProcessor);
     }}, POLL_MS);
   }}
 
-  // кнопкі кіравання
   window.__wa_start_click = async function() {{
     const api = await ensurePlayer();
     api.meta.t_click_ms = performance.now();
@@ -812,8 +825,8 @@ registerProcessor('push-player', PushPlayerProcessor);
     if (audio) {{ try {{ audio.play(); }} catch(e) {{}} }}
   }};
 
-  // запуск polling адразу пры загрузцы блока
-  pollLoop();
+  // Стартуем polling пасля загрузкі
+  startPolling();
 }})();
 </script>
 """