Create app-speech-to-text.py

app-speech-to-text.py

@@ -0,0 +1,41 @@
+import torch
+from transformers import Speech2Text2Processor, SpeechEncoderDecoderModel
+import streamlit as st
+from audio_recorder_streamlit import audio_recorder
+import numpy as np
+
+# Function to transcribe audio to text
+def transcribe_audio(audio_bytes):
+    processor = Speech2Text2Processor.from_pretrained("facebook/s2t-wav2vec2-large-en-de")
+    model = SpeechEncoderDecoderModel.from_pretrained("facebook/s2t-wav2vec2-large-en-de")
+
+    # Convert bytes to numpy array
+    audio_array = np.frombuffer(audio_bytes, dtype=np.int16)
+    
+    # Cast audio array to double precision and normalize
+    audio_tensor = torch.tensor(audio_array, dtype=torch.float32) / 32768.0
+    
+    input_values = processor(audio_tensor, return_tensors="pt", sampling_rate=16_000).input_values
+    logits = model(input_values).logits
+    predicted_ids = torch.argmax(logits, dim=-1)
+    transcription = processor.batch_decode(predicted_ids, skip_special_tokens=True)
+
+    return transcription[0]
+
+# Streamlit app
+st.title("Audio to Text Transcription")
+
+audio_bytes = audio_recorder(pause_threshold=3.0, sample_rate=16_000)
+
+if audio_bytes:
+    st.audio(audio_bytes, format="audio/wav")
+    
+    transcription = transcribe_audio(audio_bytes)
+
+    if transcription:
+        st.write("Transcription:")
+        st.write(transcription)
+    else:
+        st.write("Error: Failed to transcribe audio.")
+else:
+    st.write("No audio recorded.")