Update app.py

app.py

@@ -1,13 +1,24 @@
-import gradio as gr
 from transformers import pipeline
 from fpdf import FPDF
 import librosa
+import gradio as gr
 
-def transcribe_and_generate_pdf(audio_file):
+def transcribe_and_generate_pdf(audio_file, chunk_duration=30):
     try:
         transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-large")
-        audio, _ = librosa.load(audio_file, sr=16000)  # Resample to 16kHz
-        transcription = transcriber(audio)["text"]
+
+        audio, sr = librosa.load(audio_file, sr=None)
+
+        num_chunks = int(len(audio) / (sr * chunk_duration)) + 1
+
+        transcription = ""
+        for i in range(num_chunks):
+            start = i * sr * chunk_duration
+            end = min((i + 1) * sr * chunk_duration, len(audio))
+            chunk = audio[start:end]
+
+            chunk_transcription = transcriber(chunk)["text"]
+            transcription += chunk_transcription + " "
 
         output_pdf = "transcription.pdf"
         pdf = FPDF()
@@ -23,7 +34,7 @@ def transcribe_and_generate_pdf(audio_file):
 
 interface = gr.Interface(
     fn=transcribe_and_generate_pdf,
-    inputs=gr.Audio(type="filepath"),  # Updated to remove 'source' and use 'type="filepath"'
+    inputs=gr.Audio(type="filepath"),
     outputs=[
         gr.Textbox(label="Transcription"),
         gr.File(label="Download PDF")
@@ -33,4 +44,4 @@ interface = gr.Interface(
 )
 
 if __name__ == "__main__":
-    interface.launch() 
+    interface.launch()