Translate in Dutch

app.py

@@ -3,15 +3,18 @@ import numpy as np
 import torch
 from datasets import load_dataset
 
-from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
+from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline, AutoProcessor
 
 
 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)
+asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", chunk_length_s=30, device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
+#processor = AutoProcessor.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl")
+#model = SpeechT5ForTextToSpeech.from_pretrained("sanchit-gandhi/speecht5_tts_vox_nl").to(device)
+
 processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
 
 model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
@@ -22,12 +25,13 @@ speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze
 
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "translate"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language":"nl"})
     return outputs["text"]
 
 
 def synthesise(text):
-    inputs = processor(text=text, return_tensors="pt")
+    inputs = processor(text=text, padding='max_length', truncation=True, return_tensors="pt")
+    #inputs = processor(text=text, padding='max_length', truncation=True, return_tensors="pt")
     speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
     return speech.cpu()