Update app.py

app.py

@@ -12,6 +12,8 @@ tokenizer = AutoTokenizer.from_pretrained("alirezamsh/small100")
 model = AutoModelForSeq2SeqLM.from_pretrained("alirezamsh/small100")
 
 def translate_speech(audio, target_lang):
+    if isinstance(audio, tuple):
+        audio = audio[0]
     audio = audio.astype("float32")  # Convert audio to float32
     audio = whisper.pad_or_trim(audio, whisper_model.audio_config.sample_rate)
     mel = whisper.log_mel_spectrogram(audio).to(whisper_model.device)
@@ -34,6 +36,7 @@ def translate_speech(audio, target_lang):
     return audio_path
 
 
+
 def translate_speech_interface(audio, target_lang):
     translated_audio = translate_speech(audio, target_lang)
     translated_audio_bytes = open(translated_audio, "rb").read()