Update app.py

app.py

@@ -1,3 +1,4 @@
+ 
 import os
 import uuid
 import torch
@@ -15,24 +16,16 @@ from collections import deque
 os.environ["COQUI_TOS_AGREED"] = "1"
 
 # =========================
-# 🔥 DEVICE
+# 🔥 LOAD MODEL ONCE
 # =========================
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
-# =========================
-# 🔥 MULTI-CHANNEL MODEL POOL
-# =========================
-NUM_CHANNELS = 5
-print(f"🚀 Loading {NUM_CHANNELS} XTTS channels...")
-
-channels = []
-for i in range(NUM_CHANNELS):
-    tts_model = TTS(
-        model_name="tts_models/multilingual/multi-dataset/xtts_v2",
-        progress_bar=False
-    ).to(device)
-    channels.append({"tts": tts_model, "busy": False})
-    print(f"✅ Channel {i+1} loaded")
+print("🚀 Loading XTTS model...")
+tts = TTS(
+    model_name="tts_models/multilingual/multi-dataset/xtts_v2",
+    progress_bar=False
+).to(device)
+print("✅ Model loaded!")
 
 # =========================
 # 📁 OUTPUT DIR
@@ -41,41 +34,48 @@ OUTPUT_DIR = "outputs"
 os.makedirs(OUTPUT_DIR, exist_ok=True)
 
 # =========================
-# 🔹 REQUEST QUEUE
+# ⚡ BATCH CONFIG
 # =========================
+BATCH_SIZE = 3
+BATCH_WAIT_TIME = 1  # seconds
+
 request_queue = deque()
 
 # =========================
-# 🔥 MULTI-CHANNEL WORKER
+# 🔥 BATCH WORKER
 # =========================
-async def channel_worker(channel):
-    """Worker that processes requests assigned to this channel."""
+async def batch_worker():
+    print("🔥 Batch worker started...")
+
     while True:
         if len(request_queue) == 0:
             await asyncio.sleep(0.1)
             continue
 
-        if channel["busy"]:
-            await asyncio.sleep(0.05)
-            continue
+        # Wait to collect batch
+        await asyncio.sleep(BATCH_WAIT_TIME)
+
+        batch = []
+        while len(request_queue) > 0 and len(batch) < BATCH_SIZE:
+            batch.append(request_queue.popleft())
+
+        print(f"⚡ Processing batch of {len(batch)}")
 
-        # Pick next request
-        text, lang, audio_path, output_path, future = request_queue.popleft()
-        channel["busy"] = True
+        for item in batch:
+            text, lang, audio_path, output_path, future = item
 
-        try:
-            channel["tts"].tts_to_file(
-                text=text,
-                speaker_wav=audio_path,
-                language=lang,
-                file_path=output_path,
-                split_sentences=True
-            )
-            future.set_result(output_path)
-        except Exception as e:
-            future.set_result(str(e))
-        finally:
-            channel["busy"] = False
+            try:
+                tts.tts_to_file(
+                    text=text,
+                    speaker_wav=audio_path,
+                    language=lang,
+                    file_path=output_path,
+                    split_sentences=True
+                )
+                future.set_result(output_path)
+
+            except Exception as e:
+                future.set_result(str(e))
 
 
 # =========================
@@ -85,10 +85,7 @@ api = FastAPI()
 
 @api.on_event("startup")
 async def startup_event():
-    # Start a worker for each channel
-    for ch in channels:
-        asyncio.create_task(channel_worker(ch))
-    print(f"🔥 {NUM_CHANNELS} channel workers started!")
+    asyncio.create_task(batch_worker())
 
 
 @api.post("/clone-voice/")
@@ -107,7 +104,6 @@ async def clone_voice_api(
         loop = asyncio.get_event_loop()
         future = loop.create_future()
 
-        # Add to request queue
         request_queue.append((text, language, input_path, output_path, future))
 
         result = await future
@@ -137,6 +133,7 @@ async def clone_voice_ui(audio_path, text, language):
     future = loop.create_future()
 
     request_queue.append((text, language, audio_path, output_path, future))
+
     result = await future
 
     if isinstance(result, str) and result.endswith(".wav"):
@@ -145,8 +142,8 @@ async def clone_voice_ui(audio_path, text, language):
         return f"❌ {result}", None
 
 
-with gr.Blocks(title="XTTS Voice Cloning (Multi-Channel)") as demo:
-    gr.Markdown("# 🎤 XTTS Voice Cloning (Multi-Channel)")
+with gr.Blocks(title="XTTS Voice Cloning (Batching)") as demo:
+    gr.Markdown("# 🎤 XTTS Voice Cloning (Batch Mode)")
 
     audio_input = gr.Audio(type="filepath", label="Speaker Audio")
     text_input = gr.Textbox(label="Text")
@@ -163,6 +160,7 @@ with gr.Blocks(title="XTTS Voice Cloning (Multi-Channel)") as demo:
         outputs=[status, output_audio]
     )
 
+# ✅ FIXED QUEUE (no concurrency_count)
 demo.queue(max_size=20)
 
 # =========================