Update app.py

app.py

@@ -1,31 +1,25 @@
 import torch
+from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
 import gradio as gr
-import soundfile as sf
-from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
-from datasets import load_dataset
 
-# Load Sinhala SpeechT5 model from Hugging Face
-processor = SpeechT5Processor.from_pretrained("Ransaka/speecht5_tts_sinhala")
-model = SpeechT5ForTextToSpeech.from_pretrained("Ransaka/speecht5_tts_sinhala")
-vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
+# Load model & tokenizer
+model_name = "machinelearningzuu/sinhala-text-to-speech"
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForSeq2SeqLM.from_pretrained(model_name)
 
-# Load speaker embedding
-embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
-speaker_embedding = torch.tensor(embeddings_dataset[0]["xvector"]).unsqueeze(0)
+def tts_generate(text):
+    inputs = tokenizer(text, return_tensors="pt")
+    with torch.no_grad():
+        outputs = model.generate(**inputs)
+    audio = outputs[0].cpu().numpy()
+    return audio
 
-def tts(text):
-    inputs = processor(text=text, return_tensors="pt")
-    speech = model.generate_speech(inputs["input_ids"], speaker_embedding, vocoder=vocoder)
-    sf.write("output.wav", speech.numpy(), samplerate=16000)
-    return "output.wav"
-
-demo = gr.Interface(
-    fn=tts,
-    inputs=gr.Textbox(label="සිංහල පෙළය ඇතුළත් කරන්න", placeholder="උදා: අපිට උදව්වක් ඕනෙ."),
-    outputs=gr.Audio(type="filepath", label="ඇසෙන්නා වූ ශබ්දය"),
-    title="සිංහල Text-to-Speech",
-    description="Ransaka/speecht5_tts_sinhala මොඩලය භාවිතයෙන් සිංහල වාක්‍ය ශබ්ද බවට පරිවර්තනය කරන්න"
+iface = gr.Interface(
+    fn=tts_generate,
+    inputs=gr.Textbox(label="සිංහල පේළියක් ඇතුළත් කරන්න"),
+    outputs=gr.Audio(label="නිමැවුම් හඬ", type="numpy"),
+    title="Sinhala Text to Speech",
+    description="සිංහල වාක්‍යයක් ඇතුළත් කරන්න. AI හරහා හඬක් ලෙස output ලැබේ."
 )
 
-if __name__ == "__main__":
-    demo.launch()
+iface.launch()