INNER CODE UNIT · Python

tts

netease-youdao/EmotiVoice · demo_page.py:119

def tts(name, text, prompt, content, speaker, models):
    (style_encoder, generator, tokenizer, token2id, speaker2id)=models
    

    style_embedding = get_style_embedding(prompt, tokenizer, style_encoder)
    content_embedding = get_style_embedding(content, tokenizer, style_encoder)

    speaker = speaker2id[speaker]

    text_int = [token2id[ph] for ph in text.split()]
    
    sequence = torch.from_numpy(np.array(text_int)).to(DEVICE).long().unsqueeze(0)
    sequence_len = torch.from_numpy(np.array([len(text_int)])).to(DEVICE)
    style_embedding = torch.from_numpy(style_embedding).to(DEVICE).unsqueeze(0)
    content_embedding = torch.from_numpy(content_embedding).to(DEVICE).unsqueeze(0)
    speaker = torch.from_numpy(np.array([speaker])).to(DEVICE)

    with torch.no_grad():

View source record →