0.15

text_to_speech.py

@@ -0,0 +1,106 @@
+import requests
+import numpy as np
+import io
+import torch
+from pydub import AudioSegment
+import urllib.parse
+import os
+
+class TextToSpeech:
+    @classmethod
+    def INPUT_TYPES(cls):
+        # speakers_dir = "speakers"
+        speakers_dir = os.path.join(os.path.dirname(os.path.realpath(__file__)), "speakers")
+        speaker_options = []
+
+        for root, dirs, files in os.walk(speakers_dir):
+            for file in files:
+                if file.endswith(".wav"):
+                    rel_path = os.path.relpath(os.path.join(root, file), speakers_dir)
+                    speaker_options.append(rel_path)
+
+        # If no .wav files are found, add a default option
+        if not speaker_options:
+            speaker_options.append("No WAV files found")
+
+        return {
+            "required": {
+                "text": ("STRING", {"multiline": True}),
+                "language": (["ar", "cs", "de", "en", "es", "fr", "hi", "hu", "it", "ja", "ko", "nl", "pl", "pt", "ru", "tr", "zh-cn"], {
+                    "default": "en",
+                    "display": "dropdown",
+                    "labels": ["Arabic", "Czech", "German", "English", "Spanish", "French", "Hindi", "Hungarian", "Italian", "Japanese", "Korean", "Dutch", "Polish", "Portuguese", "Russian", "Turkish", "Chinese"]
+                }),
+                "speaker_wav": (speaker_options, {
+                    "default": speaker_options[0],
+                    "display": "dropdown"
+                }),
+            }
+        }
+
+    RETURN_TYPES = ("AUDIO",)
+    FUNCTION = "generate_audio"
+    CATEGORY = "audio"
+
+    def generate_audio(self, text, language, speaker_wav):
+        # Check if a valid speaker_wav was selected
+        if speaker_wav == "No WAV files found":
+            print("Error: No WAV files available for text-to-speech.")
+            return ({"waveform": torch.zeros(1, 1, 1, dtype=torch.float32), "sample_rate": 22050},)
+        encoded_text = urllib.parse.quote(text)  # Encode spaces and special characters
+        url = f"http://localhost:8020/tts_stream?language={language}&speaker_wav={speaker_wav}&text={encoded_text}"
+        try:
+            response = requests.get(url, stream=True)
+            response.raise_for_status()
+
+            audio_data = io.BytesIO()
+            for chunk in response.iter_content(chunk_size=8192):
+                audio_data.write(chunk)
+            
+            audio_data.seek(0)
+            return self.process_audio_data(audio_data)
+
+        except requests.RequestException as e:
+            print(f"Error generating audio: {e}")
+            return ({"waveform": torch.zeros(1, 1, 1, dtype=torch.float32), "sample_rate": 22050},)
+        except Exception as e:
+            print(f"Unexpected error: {e}")
+            return ({"waveform": torch.zeros(1, 1, 1, dtype=torch.float32), "sample_rate": 22050},)
+
+    def process_audio_data(self, audio_data):
+        try:
+            # Load MP3 data
+            audio = AudioSegment.from_mp3(audio_data)
+            
+            # Get audio properties
+            sample_rate = audio.frame_rate
+            num_channels = audio.channels
+            
+            # Convert to numpy array
+            audio_np = np.array(audio.get_array_of_samples()).astype(np.float32)
+            
+            # Normalize to [-1, 1]
+            audio_np /= np.iinfo(np.int16).max
+            
+            print(f"Raw audio data shape: {audio_np.shape}")
+            
+            # Reshape to (num_channels, num_samples)
+            if num_channels == 1:
+                audio_np = audio_np.reshape(1, -1)
+            else:
+                audio_np = audio_np.reshape(-1, num_channels).T
+            
+            # Convert to torch tensor
+            audio_tensor = torch.from_numpy(audio_np)
+            
+            print(f"Final audio tensor shape: {audio_tensor.shape}")
+            print(f"Audio data type: {audio_tensor.dtype}")
+            print(f"Audio data min: {audio_tensor.min()}, max: {audio_tensor.max()}")
+            
+            # Wrap the tensor in a list to match the expected format
+            return ({"waveform": audio_tensor.unsqueeze(0), "sample_rate": sample_rate},)
+    
+        except Exception as e:
+            print(f"Error processing audio data: {e}")
+            raise
+