cut at 60 secs

app.py

@@ -1,25 +1,24 @@
 import base64
 import os
 import time
+from io import BytesIO
 
 import banana_dev as banana
 import gradio as gr
 from loguru import logger
+from pydub import AudioSegment
 
 api_key = os.environ["BANANA_API_KEY"]
 model_key = os.environ["BANANA_MODEL_KEY"]
 password = os.environ["PASSWORD"]
 
 
-def transcribe(file=None, url=None):
-    if file:
-        with open(file, "rb") as f:
-            audio = f.read()
-            audio_b64 = base64.b64encode(audio).decode("ISO-8859-1")
+def transcribe(audio=None, url=None):
+    if audio:
+        audio_b64 = base64.b64encode(audio.export().read()).decode("ascii")
         payload = {"audio_b64": audio_b64}
     else:
         payload = {"url": url}
-    # payload["email"] = email
     response = banana.run(api_key, model_key, payload)
     print(response)
 
@@ -48,7 +47,7 @@ def run_demo(password, microphone, file_upload):
     file = microphone if microphone is not None else file_upload
 
     start = time.time()
-    transcription = transcribe(file)
+    transcription = transcribe(AudioSegment.from_file(file)[:60_000])
     logger.info(f"transcription took {time.time()-start:.3f}s")
     return "\n\n".join([seg["text"].strip() for seg in transcription["segments"]])