Update app.py

app.py

@@ -1,41 +1,31 @@
-import os
-import tempfile
+import gradio as gr
+from transformers import AutoModelForSpeechSeq2Seq, AutoProcessor
 import torch
 import soundfile as sf
+import tempfile
+
+# Load model and processor
+model_id = "jellecali8/somali_tts_model"
+processor = AutoProcessor.from_pretrained(model_id)
+model = AutoModelForSpeechSeq2Seq.from_pretrained(model_id)
 
 def tts(text):
-    try:
-        if not text.strip():
-            return None  # Qoraal madhan ha soo gelin
-        device = next(model.parameters()).device
-        inputs = tokenizer(text, return_tensors="pt").to(device)
-        custom_embedding_ = custom_embedding.to(device)
-        with torch.no_grad():
-            outputs = model(**inputs, speaker_embeddings=custom_embedding_)
-            waveform = outputs.waveform.squeeze().cpu().numpy()
-        
-        # Hubi waveform-ka: min, max, shape
-        print(f"Waveform shape: {waveform.shape}, min: {waveform.min()}, max: {waveform.max()}")
-        
-        # Normalize waveform
-        max_val = max(abs(waveform.max()), abs(waveform.min()))
-        if max_val > 0:
-            waveform = waveform / max_val
-        else:
-            print("Warning: Waveform is all zeros")
-            return None
-        
-        tmp = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
-        sf.write(tmp.name, waveform, 16000)
-        
-        # Hubi faylka la abuuray: path iyo size
-        print(f"Generated WAV file path: {tmp.name}")
-        print(f"WAV file size: {os.path.getsize(tmp.name)} bytes")
-        
-        if os.path.getsize(tmp.name) == 0:
-            print("Warning: Generated WAV file is empty")
-            return None
-        
-        return tmp.name
-    except Exception as e:
-        return f"Error: {str(e)}"
+    inputs = processor(text, return_tensors="pt")
+    with torch.no_grad():
+        outputs = model.generate(**inputs)
+    audio = outputs[0].cpu().numpy()
+
+    # Save to temporary file
+    with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as f:
+        sf.write(f.name, audio, samplerate=16000)
+        return f.name
+
+iface = gr.Interface(
+    fn=tts,
+    inputs=gr.Textbox(lines=2, label="Enter Somali Text"),
+    outputs=gr.Audio(label="Generated Speech"),
+    title="Somali TTS Demo",
+    description="Ku qor qoraalka Somali, kadib dhageyso codka."
+)
+
+iface.launch()