rvc_lib.py

import torch
import torchaudio
import os
import soundfile as sf
from ovos_tts_plugin_manager.tts import TTSFactory  # correct new way

class VoiceConverter:
    """
    RVC voice converter integrated with OpenVoiceOS TTS
    """

    def __init__(self, model_path, voice_name="en_GB-amy", device=None):
        self.device = device or ("cuda" if torch.cuda.is_available() else "cpu")
        
        if not os.path.exists(model_path):
            raise FileNotFoundError(f"Model file not found: {model_path}")
        
        # Load RVC model
        self.model = torch.load(model_path, map_location=self.device)
        self.model.eval()
        print(f"[INFO] Loaded RVC model: {model_path} on {self.device}")
        
        # Load TTS via OpenVoiceOS
        self.tts = TTSFactory.get_tts(voice_name=voice_name)
        print(f"[INFO] Loaded TTS voice: {voice_name}")

    def convert(self, input_audio_path):
        # RVC conversion
        waveform, sample_rate = torchaudio.load(input_audio_path)
        if waveform.shape[0] > 1:
            waveform = waveform.mean(dim=0, keepdim=True)  # convert to mono

        with torch.no_grad():
            converted_waveform = self.model(waveform.unsqueeze(0)).squeeze().cpu()

        output_path = input_audio_path.replace(".wav", "_converted.wav")
        sf.write(output_path, converted_waveform.numpy(), sample_rate)
        return output_path

    def synthesize(self, text, output_path="tts_output.wav"):
        # TTS generation
        audio_bytes = self.tts.get_tts(text)
        sf.write(output_path, audio_bytes, 22050)  # default Piper sample rate
        return output_path