Update app.py

app.py

@@ -14,23 +14,26 @@ asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base",
 # load text-to-speech checkpoint and speaker embeddings
 processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
 
-model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
+#model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
+#Use own TTS Model
+model = SpeechT5ForTextToSpeech.from_pretrained("jasonl1/speecht5_finetuned_voxpopuli_fi")
+
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
 
 
-def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
-    return outputs["text"]
+#def translate(audio):
+#    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
+#    return outputs["text"]
 
 
-# Added to Base to translate from Language X to any Language Y using "task": "transcribe"
+# Added to Base to translate from Language X to any Language Y ="fi" using "task": "transcribe"
 # At Inference. it should use translate(sample["audio"].copy())
 
 def translate(audio):
-    outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "es"})
+    outputs = pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "fi"})
     return outputs["text"]