take user input audi and convert to text

app.py

@@ -0,0 +1,71 @@
+import streamlit as st
+import soundfile as sf
+import librosa
+import numpy as np
+import time
+from transformers import pipeline
+from io import BytesIO
+
+# Define the models (You can replace these with any other top models supporting audio input)
+MODELS = {
+    "Whisper (English)": "openai/whisper-small.en",
+    "Whisper (Multilingual)": "openai/whisper-small",
+    "Facebook Wav2Vec2": "facebook/wav2vec2-large-960h",
+    "Google Wav2Vec2": "google/wav2vec2-large-xlsr-53",
+    "Whisper (Thai)": "openai/whisper-large"
+}
+
+# App UI
+st.title("Audio to Text Conversion")
+st.subheader("Select language and model")
+
+# Language selection
+language = st.selectbox("Choose Language", options=["English", "Thai"])
+
+# Model selection
+model_choice = st.selectbox("Choose a Model", options=list(MODELS.keys()))
+
+# Record audio
+st.subheader("Record your audio")
+audio_recorder = st.audio("")
+
+if st.button("Start Recording"):
+    # Add code here to handle audio recording via mic or upload if needed
+    st.warning("Audio recording functionality needs to be implemented")
+
+# Placeholder for conversion metrics
+if audio_recorder:
+    st.write("Recording audio metrics...")
+    
+    # Read audio file
+    audio_data, sr = librosa.load(audio_recorder, sr=None)
+
+    # Compute audio properties
+    audio_size = len(audio_data) * 2  # in bytes (16-bit PCM)
+    frame_rate = sr
+    duration = librosa.get_duration(y=audio_data, sr=sr)
+
+    # Display audio properties
+    st.write(f"Audio Size: {audio_size} bytes")
+    st.write(f"Frame Rate: {frame_rate} Hz")
+    st.write(f"Duration: {duration:.2f} seconds")
+
+    # Perform conversion using the selected model
+    st.subheader("Converting audio to text...")
+
+    start_time = time.time()
+
+    # Load the model from HuggingFace
+    model = pipeline("automatic-speech-recognition", model=MODELS[model_choice])
+
+    # Perform the conversion
+    audio_bytes = BytesIO(sf.write("temp.wav", audio_data, sr))
+    result = model(audio_bytes)
+
+    end_time = time.time()
+
+    # Display results
+    st.write("Transcription:", result['text'])
+    st.write(f"Conversion took {end_time - start_time:.2f} seconds")
+
+# Provide placeholder for actual audio recording functionality if necessary.

requirements.txt

@@ -0,0 +1,4 @@
+streamlit
+transformers
+librosa
+soundfile