doby4u/chattts

1import os2import random3import argparse4 5import torch6import gradio as gr7import numpy as np8 9import ChatTTS10 11print("loading ChatTTS model...")12chat = ChatTTS.Chat()13chat.load_models()14 15def generate_seed():16    new_seed = random.randint(1, 100000000)17    return {18        "__type__": "update",19        "value": new_seed20        }21 22 23def generate_audio(text, temperature, top_P, top_K, audio_seed_input, text_seed_input, refine_text_flag):24 25    torch.manual_seed(audio_seed_input)26    rand_spk = torch.randn(768)27    params_infer_code = {28        'spk_emb': rand_spk, 29        'temperature': temperature,30        'top_P': top_P,31        'top_K': top_K,32        }33    params_refine_text = {'prompt': '[oral_2][laugh_0][break_6]'}34    35    torch.manual_seed(text_seed_input)36 37    if refine_text_flag:38        text = chat.infer(text, 39                          skip_refine_text=False,40                          refine_text_only=True,41                          params_refine_text=params_refine_text,42                          params_infer_code=params_infer_code43                          )44    45    wav = chat.infer(text, 46                     skip_refine_text=True, 47                     params_refine_text=params_refine_text, 48                     params_infer_code=params_infer_code49                     )50    51    audio_data = np.array(wav[0]).flatten()52    sample_rate = 2400053    text_data = text[0] if isinstance(text, list) else text54 55    return [(sample_rate, audio_data), text_data]56 57 58def main():59 60 61    with gr.Blocks() as demo:62        gr.Markdown("# ChatTTS Webui")63        gr.Markdown("ChatTTS Model: [2noise/ChatTTS](https://github.com/2noise/ChatTTS)")64 65        default_text = "四川美食确实以辣闻名，但也有不辣的选择。比如甜水面、赖汤圆、蛋烘糕、叶儿粑等，这些小吃口味温和，甜而不腻，也很受欢迎。"        66        text_input = gr.Textbox(label="Input Text", lines=4, placeholder="Please Input Text...", value=default_text)67 68        with gr.Row():69            refine_text_checkbox = gr.Checkbox(label="Refine text", value=True)70            temperature_slider = gr.Slider(minimum=0.00001, maximum=1.0, step=0.00001, value=0.3, label="Audio temperature")71            top_p_slider = gr.Slider(minimum=0.1, maximum=0.9, step=0.05, value=0.7, label="top_P")72            top_k_slider = gr.Slider(minimum=1, maximum=20, step=1, value=20, label="top_K")73 74        with gr.Row():75            audio_seed_input = gr.Number(value=42, label="Audio Seed")76            generate_audio_seed = gr.Button("\U0001F3B2")77            text_seed_input = gr.Number(value=42, label="Text Seed")78            generate_text_seed = gr.Button("\U0001F3B2")79 80        generate_button = gr.Button("Generate")81        82        text_output = gr.Textbox(label="Output Text", interactive=False)83        audio_output = gr.Audio(label="Output Audio")84 85        generate_audio_seed.click(generate_seed, 86                                  inputs=[], 87                                  outputs=audio_seed_input)88        89        generate_text_seed.click(generate_seed, 90                                 inputs=[], 91                                 outputs=text_seed_input)92        93        generate_button.click(generate_audio, 94                              inputs=[text_input, temperature_slider, top_p_slider, top_k_slider, audio_seed_input, text_seed_input, refine_text_checkbox], 95                              outputs=[audio_output, text_output])96 97    parser = argparse.ArgumentParser(description='ChatTTS demo Launch')98    parser.add_argument('--server_name', type=str, default='0.0.0.0', help='Server name')99    parser.add_argument('--server_port', type=int, default=8080, help='Server port')100    args = parser.parse_args()101 102    # demo.launch(server_name=args.server_name, server_port=args.server_port, inbrowser=True)103    demo.launch()104 105 106if __name__ == '__main__':107    main()