Update app.py

app.py

@@ -3,7 +3,7 @@ import numpy as np
 import torch
 from datasets import load_dataset
 
-from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline, WhisperProcessor, BarkModel, BarkProcessor
+from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline, WhisperProcessor
 
 
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -19,8 +19,12 @@ asr_pipe = pipeline(
     device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
-processor = BarkProcessor.from_pretrained("suno/bark-small")
-model = BarkModel.from_pretrained("suno/bark-small").to(device)
+processor = SpeechT5Processor.from_pretrained("Sandiago21/speecht5_finetuned_facebook_voxpopuli_spanish")
+model = SpeechT5ForTextToSpeech.from_pretrained("Sandiago21/speecht5_finetuned_facebook_voxpopuli_spanish").to(device)
+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):
@@ -29,17 +33,15 @@ def translate(audio):
 
 
 def synthesise(text):
-    inputs = processor(text, voice_preset="v2/es_speaker_3")
-    speech = model.generate(**inputs).cpu()
-    return speech
+    inputs = processor(text=text, return_tensors="pt")
+    speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
+    return speech.cpu()
 
 
 def speech_to_speech_translation(audio):
     translated_text = translate(audio)
     synthesised_speech = synthesise(translated_text)
-    synthesised_speech = synthesised_speech.numpy()
-    synthesised_speech = np.clip(synthesised_speech, -1.0, 1.0)
-    synthesised_speech = (synthesised_speech * 32767).astype(np.int16)
+    synthesised_speech = (synthesised_speech.numpy() * 32767).astype(np.int16)
     return 16000, synthesised_speech
 
 
@@ -73,4 +75,4 @@ file_translate = gr.Interface(
 with demo:
     gr.TabbedInterface([mic_translate, file_translate], ["Microphone", "Audio File"])
 
-demo.launch()
+demo.launch()