Update app.py

app.py

@@ -9,12 +9,14 @@ from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Proce
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
 # load speech translation checkpoint
+
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
 processor = SpeechT5Processor.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
 
 model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl").to(device)
+
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
@@ -22,6 +24,8 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
 
 
 def translate(audio):
+    print(audio)
+    audio = librosa.resample(audio, orig_sr=22050, target_sr=16000)
     outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"language": "nl","task": "transcribe"})
     return outputs["text"]
 
@@ -69,4 +73,4 @@ file_translate = gr.Interface(
 with demo:
     gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
 
-demo.launch()
+demo.launch(share=True)