Update app.py

app.py

@@ -1,21 +1,17 @@
-from fastapi import FastAPI, Request
-from fastapi.responses import FileResponse
+import gradio as gr
 import torch
 import numpy as np
 import scipy.io.wavfile
 from transformers import VitsModel, AutoTokenizer
 import re
 
-app = FastAPI()
-
-# Load your model
+# Load fine-tuned model from Hugging Face Hub or local path
 model = VitsModel.from_pretrained("Somali-tts/somali_tts_model")
 tokenizer = AutoTokenizer.from_pretrained("saleolow/somali-mms-tts")
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 model.to(device)
 model.eval()
 
-# Number conversion (keep your existing number_words + number_to_words)
 number_words = {
     0: "eber", 1: "koow", 2: "labo", 3: "seddex", 4: "afar", 5: "shan",
     6: "lix", 7: "todobo", 8: "sideed", 9: "sagaal", 10: "toban",
@@ -79,13 +75,20 @@ def normalize_text(text):
     text = text.replace("ZamZam", "SamSam")
     return text
 
-@app.post("/tts")
-async def tts(request: Request):
-    data = await request.json()
-    text = normalize_text(data["text"])
+def tts(text):
+    text = normalize_text(text)
     inputs = tokenizer(text, return_tensors="pt").to(device)
     with torch.no_grad():
         waveform = model(**inputs).waveform.squeeze().cpu().numpy()
     filename = "output.wav"
     scipy.io.wavfile.write(filename, rate=model.config.sampling_rate, data=(waveform * 32767).astype(np.int16))
-    return FileResponse(filename, media_type="audio/wav")
+    return filename
+
+gr.Interface(
+    fn=tts,
+    inputs=gr.Textbox(label="Geli qoraal Soomaali ah"),
+    outputs=gr.Audio(label="Codka TTS"),
+    title="Somali TTS",
+    description="Ku qor qoraal Soomaaliyeed si aad u maqasho cod dabiici ah.",
+).launch()
+