Create app.py

app.py

@@ -0,0 +1,60 @@
+from fastapi import FastAPI, Form, HTTPException
+from fastapi.responses import FileResponse
+from kokoro import KPipeline
+import torch
+import soundfile as sf
+import numpy as np
+import io
+import base64
+
+app = FastAPI(title="Kokoro TTS API", description="API لتحويل النص إلى كلام باستخدام Kokoro TTS")
+
+# Automatically use GPU if available, otherwise use CPU
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+print(f"Using device: {device}")
+
+# Initialize KPipeline globally when the app starts
+try:
+    pipeline = KPipeline(lang_code='a', device=device) # Assuming 'a' is still the correct lang_code
+except Exception as e:
+    print(f"Error initializing KPipeline: {e}")
+    pipeline = None # Handle initialization failure
+
+@app.post("/tts/")
+async def text_to_speech(text: str = Form(...)):
+    """
+    تحويل النص إلى كلام باستخدام نموذج Kokoro TTS.
+
+    - **text**: النص المراد تحويله إلى كلام (إلزامي).
+    """
+    if pipeline is None:
+        raise HTTPException(status_code=500, detail="Kokoro TTS Pipeline لم يتم تهيئته بشكل صحيح.")
+
+    if not text:
+        raise HTTPException(status_code=400, detail="يجب توفير نص لتحويله إلى كلام.")
+
+    try:
+        generator = pipeline(text, voice='af_heart', speed=1, split_pattern=None)
+        audio_segments = []
+        for _, _, audio in generator:
+            audio_segments.append(audio)
+        full_audio = np.concatenate(audio_segments, axis=0)
+
+        # Save to BytesIO buffer in WAV format
+        wav_buffer = io.BytesIO()
+        sf.write(wav_buffer, full_audio, 24000, format='WAV')
+        wav_bytes = wav_buffer.getvalue()
+
+        # Encode to base64 for embedding in JSON response (optional, can also return FileResponse)
+        # audio_base64 = base64.b64encode(wav_bytes).decode('utf-8')
+        # return {"audio_base64": audio_base64}
+
+        # Save to a temporary file and return FileResponse (more efficient for larger audio files)
+        with open("output.wav", "wb") as wav_file:
+            wav_file.write(wav_bytes)
+        return FileResponse("output.wav", media_type="audio/wav", filename="output.wav")
+
+
+    except Exception as e:
+        print(f"Error during TTS processing: {e}")
+        raise HTTPException(status_code=500, detail=f"حدث خطأ أثناء معالجة النص إلى كلام: {e}")