Update app.py

app.py

@@ -1,4 +1,5 @@
 import streamlit as st
+import whisper
 from transformers import pipeline
 import speech_recognition as sr
 from gtts import gTTS
@@ -6,93 +7,72 @@ import tempfile
 import os
 import base64
 
-# Initialize the translation pipeline
+# Load Whisper model for Speech-to-Text
 @st.cache_resource
-def load_translation_pipeline(Helsinki-NLP/opus-mt):
-    return pipeline("translation", model=Helsinki-NLP/opus-mt)
+def load_whisper_model():
+    return whisper.load_model("base")  # You can use other models like "large" for better accuracy
+
+# Load translation model from Hugging Face
+@st.cache_resource
+def load_translation_model(input_lang, output_lang):
+    model_name = f"Helsinki-NLP/opus-mt-{input_lang}-{output_lang}"
+    return pipeline("translation", model=model_name)
 
 # Initialize speech recognition
 recognizer = sr.Recognizer()
 
-def translate_text(input_text, model):
-    return model(input_text)[0]["translation_text"]
-
 def speech_to_text(audio_file):
     with sr.AudioFile(audio_file) as source:
         audio_data = recognizer.record(source)
         return recognizer.recognize_google(audio_data)
 
+def translate_text(input_text, translation_pipeline):
+    return translation_pipeline(input_text)[0]["translation_text"]
+
 def text_to_speech(text, language):
     tts = gTTS(text=text, lang=language)
     temp_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp3")
     tts.save(temp_file.name)
     return temp_file.name
 
-# Streamlit app
-st.title("Real-Time Language Translator")
-st.write("Translate voice and text between multiple languages in real-time!")
+# Streamlit UI
+st.title("Voice-to-Voice Translator")
 
-# Language selection
+# Sidebar for user input
 st.sidebar.header("Settings")
-input_lang = st.sidebar.selectbox("Select Input Language", ["English", "French", "Spanish", "German", "Hindi"])
-output_lang = st.sidebar.selectbox("Select Output Language", ["English", "French", "Spanish", "German", "Hindi"])
-
-# Language codes mapping
-lang_codes = {
-    "English": "en",
-    "French": "fr",
-    "Spanish": "es",
-    "German": "de",
-    "Hindi": "hi"
-}
+input_lang = st.sidebar.selectbox("Select Input Language", ["en", "fr", "es", "de", "hi"])
+output_lang = st.sidebar.selectbox("Select Output Language", ["en", "fr", "es", "de", "hi"])
 
-input_code = lang_codes[input_lang]
-output_code = lang_codes[output_lang]
-
-# Model selection
-model_name = f"Helsinki-NLP/opus-mt-{input_code}-{output_code}"
-translation_pipeline = load_translation_pipeline(model_name)
+# Initialize models
+whisper_model = load_whisper_model()
+translation_pipeline = load_translation_model(input_lang, output_lang)
 
 # Input options
-st.header("Input Options")
-input_mode = st.radio("Choose Input Mode:", ["Text", "Voice"])
-
-if input_mode == "Text":
-    input_text = st.text_area(f"Enter text in {input_lang}:")
-    if st.button("Translate"):
-        if input_text.strip():
-            translated_text = translate_text(input_text, translation_pipeline)
-            st.success(f"Translated Text in {output_lang}: {translated_text}")
-
-            # Option to download translation as audio
-            if st.checkbox("Play Translated Audio"):
-                audio_file = text_to_speech(translated_text, output_code)
-                audio_bytes = open(audio_file, "rb").read()
-                st.audio(audio_bytes, format="audio/mp3")
+st.header("Input your voice or text")
+
+# Option to upload an audio file
+audio_file = st.file_uploader("Upload an audio file (WAV format)", type=["wav"])
+
+if audio_file is not None:
+    # Convert speech to text
+    input_text = speech_to_text(audio_file)
+    st.write(f"Recognized Text: {input_text}")
+    
+    # Translate the text
+    translated_text = translate_text(input_text, translation_pipeline)
+    st.success(f"Translated Text: {translated_text}")
+
+    # Convert translated text to speech
+    output_audio = text_to_speech(translated_text, output_lang)
+    audio_bytes = open(output_audio, "rb").read()
+    
+    # Play the audio
+    st.audio(audio_bytes, format="audio/mp3")
+    
+    # Provide download link for audio
+    b64 = base64.b64encode(audio_bytes).decode()
+    href = f'<a href="data:audio/mp3;base64,{b64}" download="translated_audio.mp3">Download Translated Audio</a>'
+    st.markdown(href, unsafe_allow_html=True)
 
-                # Provide download link
-                b64 = base64.b64encode(audio_bytes).decode()
-                href = f'<a href="data:audio/mp3;base64,{b64}" download="translation.mp3">Download Translated Audio</a>'
-                st.markdown(href, unsafe_allow_html=True)
 else:
-    audio_file = st.file_uploader("Upload an audio file (WAV format)", type=["wav"])
-    if audio_file is not None:
-        if st.button("Translate"):
-            try:
-                input_text = speech_to_text(audio_file)
-                st.write(f"Recognized Text in {input_lang}: {input_text}")
-                translated_text = translate_text(input_text, translation_pipeline)
-                st.success(f"Translated Text in {output_lang}: {translated_text}")
-
-                # Option to download translation as audio
-                if st.checkbox("Play Translated Audio"):
-                    audio_file = text_to_speech(translated_text, output_code)
-                    audio_bytes = open(audio_file, "rb").read()
-                    st.audio(audio_bytes, format="audio/mp3")
-
-                    # Provide download link
-                    b64 = base64.b64encode(audio_bytes).decode()
-                    href = f'<a href="data:audio/mp3;base64,{b64}" download="translation.mp3">Download Translated Audio</a>'
-                    st.markdown(href, unsafe_allow_html=True)
-            except Exception as e:
-                st.error(f"Error: {e}")
+    st.write("Please upload an audio file to get started.")