update tts

app.py

@@ -7,6 +7,7 @@ import numpy as np
 from nemo.collections.tts.models import FastPitchModel
 from nemo.collections.tts.models import HifiGanModel
 from nemo.collections.tts.models import MixerTTSModel
+from bark import SAMPLE_RATE, generate_audio, preload_models
 
 from transformers import pipeline
 
@@ -25,18 +26,11 @@ def greet(name):
 
 def generate_tts(text: str, speaker: int = 0):
     sr = 44100
-    # parsed = spec_generator.parse(text)
-    # spectrogram = spec_generator.generate_spectrogram(tokens=parsed, speaker=speaker)
-    # audio = voc_model.convert_spectrogram_to_audio(spec=spectrogram)
-    output = pipe(text)
-
-    # with tempfile.NamedTemporaryFile(suffix=".wav", delete=False) as fp:
-    #     torchaudio.save(fp.name, audio.to('cpu'), sample_rate=sr)
-
-    #     return fp.name
-    audio_arr = (output["audio"] * 32767).astype(np.int16)
-
-    return (output["sampling_rate"], audio_arr)
+    parsed = spec_generator.parse(text)
+    spectrogram = spec_generator.generate_spectrogram(tokens=parsed, speaker=speaker)
+    audio = voc_model.convert_spectrogram_to_audio(spec=spectrogram)
+    
+    return (sr, audio.squeeze(0).cpu().numpy())
 
 def run():
     demo = gr.Interface(