syclone.py

# 终端执行命令：
# cd /Volumes/AI/AI/GPT-SoVITS && conda activate GPTSoVits && ./go-api_v2.py.command

# API接口代码为：
# http://127.0.0.1:9880/tts?text=输入需要语音合成的内容。&text_lang=zh&ref_audio_path=./output/slicer_opt/hao.wav&prompt_lang=zh&prompt_text=爸爸，胡老师还以为爸爸没有来，然后我就去哪个墙壁里面躲着。&text_split_method=cut5&batch_size=1&media_type=wav&streaming_mode=true
# http://127.0.0.1:9880/tts?text=输入需要语音合成的内容。&text_lang=zh&ref_audio_path=./output/slicer_opt/yanbo109.wav&prompt_lang=zh&prompt_text=北宋诗人苏轼曾说，寄浮游于天地，渺沧海之遗粟。&text_split_method=cut5&batch_size=1&media_type=wav&streaming_mode=true


# YB接口
# http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_weights/yanbo109-e15.ckpt

# WQ接口
# http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_weights/weiqin109-e15.ckpt

# V1接口
# http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt
# http://127.0.0.1:9880/set_sovits_weights?weights_path=GPT_SoVITS/pretrained_models/s2G488k.pth

# V2接口
# http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt
# http://127.0.0.1:9880/set_sovits_weights?weights_path=GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth

# V3接口
# http://127.0.0.1:9880/set_gpt_weights?weights_path=GPT_SoVITS/pretrained_models/s1v3.ckpt
# http://127.0.0.1:9880/set_sovits_weights?weights_path=GPT_SoVITS/pretrained_models/s2Gv3.pth

# 重启
# http://127.0.0.1:9880/control?command=restart


import gradio as gr
import requests
import os
from pathlib import Path
import shutil
from datetime import datetime

# 设置住文件目录
sydir = "/Volumes/AI/AI/GPT-SoVITS/output/slicer_opt"

def copy_file_to_destination(uploaded_file, destination_dir):
    # 确保目标目录存在
    os.makedirs(destination_dir, exist_ok=True)
    
    # 构建目标文件路径
    dest_path = os.path.join(destination_dir, os.path.basename(uploaded_file))
    
    # 复制文件
    shutil.copy2(uploaded_file, dest_path)
    return dest_path

def generate_audio(text, ref_audio, prompt_text):
    if not text:
        return "请输入要合成的文本！", None
    
    try:
        # 构建API请求URL
        base_url = "http://127.0.0.1:9880/tts"
        
        # 修复目标目录路径
        # dest_dir = os.path.join("F:", "SOFT", "GPT-SoVITS-v3-20250212", "output", "slicer_opt")
        dest_dir = sydir
        
        # 如果有上传新的音频文件，复制到目标目录
        if ref_audio:
            copy_file_to_destination(ref_audio, dest_dir)
            ref_path = f"./output/slicer_opt/{os.path.basename(ref_audio)}"
        else:
            ref_path = "./output/slicer_opt/hao.wav"
            
        params = {
            "text": text,
            "text_lang": "zh",
            "ref_audio_path": ref_path,
            "prompt_lang": "zh",
            "prompt_text": prompt_text or "爸爸，胡老师还以为爸爸没有来，然后我就去哪个墙壁里面躲着。",
            "text_split_method": "cut5",
            "batch_size": "1",
            "media_type": "wav",
            "streaming_mode": "true"
        }
        
        # 如果有上传新的音频文件，复制到目标目录
        if ref_audio:
            dest_dir = sydir
            copy_file_to_destination(ref_audio, dest_dir)
        
        # 发送API请求
        response = requests.get(base_url, params=params, timeout=30)
        
        if response.status_code == 200:
            # 保存音频文件
            # audio_file = "output.wav"
            # 提取文件名（带扩展名）
            filename = os.path.basename(ref_path)  # "hao.wav"
            # 分割文件名和扩展名
            basename, ext = os.path.splitext(filename)  # basename="hao", ext=".wav"
            # 生成日期时间戳（格式示例：20230306-111514）
            timestamp = datetime.now().strftime("%Y%m%d-%H%M%S")
            # 组合新文件名
            # audio_file = f"{basename}-{timestamp}{ext}"  # "hao-20230306-111514.wav"
            
            # 定义目标目录和文件名
            target_dir = "声音生成"  # 子目录名称
            audio_file = os.path.join(target_dir, f"{basename}-{timestamp}{ext}")  # 路径拼接

            # 确保目标目录存在（如果不存在则创建）
            os.makedirs(target_dir, exist_ok=True)

            with open(audio_file, "wb") as f:
                f.write(response.content)
            return f"{audio_file} 音频生成成功！", audio_file
        else:
            return f"生成失败：HTTP状态码{response.status_code}", None
            
    except requests.Timeout:
        return "请求超时，请检查服务是否正常运行", None
    except requests.ConnectionError:
        return "连接失败，请检查服务是否启动", None
    except Exception as e:
        return f"发生错误：{str(e)}", None


def switch_model(model_name):
    base_url = "http://127.0.0.1:9880"
    try:
        if model_name == "YB接口":
            response = requests.get(f"{base_url}/set_gpt_weights", params={
                "weights_path": "GPT_weights/yanbo109-e15.ckpt"
            })
            return f"YB模型切换状态: {response.status_code}", f"{base_url}/set_gpt_weights?weights_path=GPT_weights/yanbo109-e15.ckpt"
        
        elif model_name == "WQ接口":
            response = requests.get(f"{base_url}/set_gpt_weights", params={
                "weights_path": "GPT_weights/weiqin109-e15.ckpt"
            })
            return f"WQ模型切换状态: {response.status_code}", f"{base_url}/set_gpt_weights?weights_path=GPT_weights/weiqin109-e15.ckpt"
        
        elif model_name in ["V1接口", "V2接口", "V3接口","F2024接口","蔡徐坤pro"]:
            paths = {
                "V1接口": {
                    "gpt": "GPT_SoVITS/pretrained_models/s1bert25hz-2kh-longer-epoch=68e-step=50232.ckpt",
                    "sovits": "GPT_SoVITS/pretrained_models/s2G488k.pth"
                },
                "V2接口": {
                    "gpt": "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s1bert25hz-5kh-longer-epoch=12-step=369668.ckpt",
                    "sovits": "GPT_SoVITS/pretrained_models/gsv-v2final-pretrained/s2G2333k.pth"
                },
                "V3接口": {
                    "gpt": "GPT_SoVITS/pretrained_models/s1v3.ckpt",
                    "sovits": "GPT_SoVITS/pretrained_models/s2Gv3.pth"
                },
                "F2024接口": {
                    "gpt": "GPT_weights_v2/f1027-e15.ckpt",
                    "sovits": "SoVITS_weights_v2/f1027_e8_s208.pth"
                },
                "蔡徐坤pro": {
                    "gpt": "GPT_weights_v2/蔡徐坤pro-e15.ckpt",
                    "sovits": "SoVITS_weights_v2/蔡徐坤pro_e12_s660.pth"
                }
            }
            
            gpt_response = requests.get(f"{base_url}/set_gpt_weights", params={
                "weights_path": paths[model_name]["gpt"]
            })
            sovits_response = requests.get(f"{base_url}/set_sovits_weights", params={
                "weights_path": paths[model_name]["sovits"]
            })
            
            status = f"{model_name}切换状态:\nGPT: {gpt_response.status_code}\nSoVITS: {sovits_response.status_code}"
            urls = f"GPT URL: {base_url}/set_gpt_weights?weights_path={paths[model_name]['gpt']}\nSoVITS URL: {base_url}/set_sovits_weights?weights_path={paths[model_name]['sovits']}"
            return status, urls
            
    except Exception as e:
        return f"切换失败：{str(e)}", "无法获取URL"

# 用于获取 ./output/slicer_opt/ 目录下的所有 WAV 文件
def get_wav_files():
    wav_dir = sydir
    wav_files = [f for f in os.listdir(wav_dir) if f.endswith('.wav')]
    return wav_files

# 用于从文字.txt文件中读取对应的提示文本
def get_prompt_text(wav_filename):
    try:
        txt_path = os.path.join(sydir, "文字.txt")  # 使用 os.path.join 来正确拼接路径
        if not os.path.exists(txt_path):
            # print(f"文件不存在: {txt_path}")
            return "提示：文字.txt 文件不存在，将使用默认提示文本"
            
        with open(txt_path, 'r', encoding='utf-8') as f:
            lines = f.readlines()
        
        # 移除文件名中的.wav后缀
        wav_name = wav_filename.replace('.wav', '')
        
        # 构建音频名和文本的映射
        text_map = {}
        current_key = None
        
        for line in lines:
            line = line.strip()
            if not line:  # 跳过空行
                continue
            if line == '-----':  # 跳过分隔符
                continue
            if current_key is None:
                current_key = line
            else:
                text_map[current_key] = line
                current_key = None
        
        # 查找对应的文本
        if wav_name in text_map:
            print(f"找到匹配: {wav_name} -> {text_map[wav_name]}")
            return text_map[wav_name]
        
        print(f"未找到匹配的文本，文件名: {wav_name}")
        print(f"当前映射表: {text_map}")
        # return ""
        return text_map.get(wav_name, "未找到对应的提示文本，将使用默认提示文本")
        
    except UnicodeDecodeError:
        return "文件编码错误，请确保文字.txt为UTF-8编码"
    except Exception as e:
        return f"读取提示文本错误：{str(e)}"

# 用于处理声音选择后的自动更新
def update_audio_and_prompt(wav_filename):
    if not wav_filename:
        return None, ""
    
    wav_path = os.path.join(sydir, wav_filename)
    prompt = get_prompt_text(wav_filename)
    return wav_path, prompt

# 创建界面
with gr.Blocks(title="BOZO 专用声音生成器") as demo:
    with gr.Column():
        gr.Markdown("# BOZO 专用声音生成器")
        
        # 文本输入区域移到最上方
        text_input = gr.Textbox(
            label="请输入要合成的文本",
            placeholder="输入需要语音合成的内容",
            lines=5
        )
        
        # 声音选择区域
        with gr.Row():
            voice_dropdown = gr.Dropdown(
                choices=get_wav_files(),
                label="选择声音",
                value=None
            )
        
        # 音频上传和提示文本并排显示
        with gr.Row():
            audio_input = gr.Audio(
                label="上传参考音频文件（可选）",
                type="filepath",
                format="wav"
            )
            prompt_input = gr.Textbox(
                label="提示文本",
                value="爸爸，胡老师还以为爸爸没有来，然后我就去哪个墙壁里面躲着。",
                placeholder="输入提示文本",
                lines=7
            )
        
        # 生成按钮
        generate_btn = gr.Button("生成音频")
        
        # 状态信息和音频输出并排显示
        with gr.Row():
            output_text = gr.Textbox(label="状态信息")
            output_audio = gr.Audio(label="生成的音频")
        
        # 模型选择区域
        with gr.Row():
            model_dropdown = gr.Dropdown(
                choices=["YB接口", "WQ接口", "V1接口", "V2接口", "V3接口", "F2024接口", "蔡徐坤pro"],
                label="选择模型接口",
                value="YB接口"
            )
            switch_btn = gr.Button("切换模型")
        
        with gr.Row():
            model_status = gr.Textbox(label="模型切换状态", lines=3)
            model_url = gr.Textbox(label="接口URL", lines=3)
        
        # 添加声音选择的响应函数
        voice_dropdown.change(
            fn=update_audio_and_prompt,
            inputs=[voice_dropdown],
            outputs=[audio_input, prompt_input]
        )
        generate_btn.click(
            fn=generate_audio,
            inputs=[text_input, audio_input, prompt_input],
            outputs=[output_text, output_audio]
        )
        
        switch_btn.click(
            fn=switch_model,
            inputs=[model_dropdown],
            outputs=[model_status, model_url]
        )
# 启动应用
if __name__ == "__main__":
    demo.launch(inbrowser=True)