Update app.py

app.py

@@ -1,35 +1,46 @@
 import gradio as gr
-import torch
-from transformers import pipeline
+import whisper
+from transformers import MarianMTModel, MarianTokenizer
+from TTS.api import TTS
 
 # Load models
-speech_to_text = pipeline("automatic-speech-recognition", model="openai/whisper-small")
-translation = pipeline("translation", model="Helsinki-NLP/opus-mt-hi-en")
-text_to_speech = pipeline("text-to-speech", model="espnet/kan-bayashi_ljspeech_vits")
-
-def translate_speech(audio):
-    # Step 1: Hindi speech to Hindi text
-    hindi_text = speech_to_text(audio)["text"]
-    
-    # Step 2: Hindi text to English translation
-    english_text = translation(hindi_text)[0]["translation_text"]
-    
-    # Step 3: English text to speech
-    english_speech = text_to_speech(english_text)["audio"]
-    
-    return hindi_text, english_text, (48000, english_speech)
+print("Loading Whisper (for Hindi STT)...")
+whisper_model = whisper.load_model("small")
+
+print("Loading MarianMT (for Hindi to English)...")
+translator_tokenizer = MarianTokenizer.from_pretrained("Helsinki-NLP/opus-mt-hi-en")
+translator_model = MarianMTModel.from_pretrained("Helsinki-NLP/opus-mt-hi-en")
+
+print("Loading TTS model (Tacotron2 + HiFi-GAN)...")
+tts = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False, gpu=False)
+
+# Pipeline function
+def hindi_speech_to_english_audio(audio):
+    # Step 1: Hindi Speech to Text
+    result = whisper_model.transcribe(audio, language="hi")
+    hindi_text = result["text"]
+
+    # Step 2: Hindi to English Translation
+    inputs = translator_tokenizer(hindi_text, return_tensors="pt", padding=True)
+    translated_tokens = translator_model.generate(**inputs)
+    english_text = translator_tokenizer.decode(translated_tokens[0], skip_special_tokens=True)
+
+    # Step 3: English Text to Speech
+    tts.tts_to_file(text=english_text, file_path="output.wav")
+
+    return english_text, "output.wav"
 
 # Gradio Interface
-demo = gr.Interface(
-    fn=translate_speech,
-    inputs=gr.Audio(source="microphone", type="filepath"),
+interface = gr.Interface(
+    fn=hindi_speech_to_english_audio,
+    inputs=gr.Audio(sources=["microphone"], type="filepath", label="🎙️ Speak in Hindi"),
     outputs=[
-        gr.Textbox(label="Hindi Text"),
-        gr.Textbox(label="English Translation"),
-        gr.Audio(label="English Audio")
+        gr.Textbox(label="🔤 Translated English Text"),
+        gr.Audio(type="filepath", label="🗣️ English Audio Output")
     ],
-    title="Hindi to English Speech Translator",
-    description="🎤 Speak in Hindi and get the English translation spoken aloud"
+    title="Hindi Speech to English Audio Translator",
+    description="🎧 Speak in Hindi and hear it back in English!",
 )
 
-demo.launch()
+# Run app
+interface.launch()