feat: add bark long form speech synthesize

app.py

@@ -104,7 +104,7 @@ def callback(indata, frames, time, status):
 
                 with console.status("Generating response...", spinner="earth"):
                     response = get_llm_response(text)
-                    sample_rate, audio_array = tts.synthesize(response)
+                    sample_rate, audio_array = tts.long_form_synthesize(response)
 
                 console.print(f"[cyan]Assistant: {response}")
                 sd.play(audio_array, sample_rate)

pyproject.toml

@@ -17,6 +17,7 @@ speechrecognition = "^3.10.1"
 pyaudio = "^0.2.14"
 langchain = "^0.1.13"
 vocos = "^0.1.0"
+nltk = "^3.8.1"
 
 
 [tool.poetry.group.dev.dependencies]

tts.py

@@ -1,5 +1,7 @@
+import nltk
 import torch
 import warnings
+import numpy as np
 from transformers import AutoProcessor, BarkModel
 
 warnings.filterwarnings(
@@ -15,7 +17,7 @@ def __init__(self, device: str = "cuda" if torch.cuda.is_available() else "cpu")
         self.model = BarkModel.from_pretrained("suno/bark-small")
         self.model.to(self.device)
 
-    def synthesize(self, text: str, voice_preset: str = "v2/en_speaker_1"):
+    def synthesize(self, text: str, voice_preset: str):
         inputs = self.processor(text, voice_preset=voice_preset, return_tensors="pt")
         inputs = {k: v.to(self.device) for k, v in inputs.items()}
 
@@ -25,3 +27,14 @@ def synthesize(self, text: str, voice_preset: str = "v2/en_speaker_1"):
         audio_array = audio_array.cpu().numpy().squeeze()
         sample_rate = self.model.generation_config.sample_rate
         return sample_rate, audio_array
+
+    def long_form_synthesize(self, text: str, voice_preset: str = "v2/en_speaker_1"):
+        pieces = []
+        sentences = nltk.sent_tokenize(text)
+        silence = np.zeros(int(0.25 * self.model.generation_config.sample_rate))
+
+        for sent in sentences:
+            sample_rate, audio_array = self.synthesize(sent, voice_preset)
+            pieces += [audio_array, silence.copy()]
+
+        return self.model.generation_config.sample_rate, np.concatenate(pieces)