当前位置：首页 > news >正文

开源语音合成模型SparkTTS使用

news 来源：原创 2025/5/17 11:42:01

一、环境配置

git clone https://github.com/SparkAudio/Spark-TTS.git

pip install -r requirements.txt

二、模型下载

从modelscope进行下载，pip install modelscope

创建一个download.py

import torchfrom modelscope import snapshot_downloadsnapshot_download("SparkAudio/Spark-TTS-0.5B", local_dir="SparkAudio/Spark-TTS-0.5B")

执行python download.py进行下载。

三、使用

在example文件夹下能够看到infer脚本文件，

#!/bin/bash# Copyright (c) 2025 SparkAudio
#               2025 Xinsheng Wang (w.xinshawn@gmail.com)
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.# Get the absolute path of the script's directory
script_dir=$(dirname "$(realpath "$0")")# Get the root directory
root_dir=$(dirname "$script_dir")# Set default parameters
device=0
save_dir='example/results'
model_dir="pretrained_models/Spark-TTS-0.5B"
text="身临其境，换新体验。塑造开源语音合成新范式，让智能语音更自然。"
prompt_text="吃燕窝就选燕之屋，本节目由26年专注高品质燕窝的燕之屋冠名播出。豆奶牛奶换着喝，营养更均衡，本节目由豆本豆豆奶特约播出。"
prompt_speech_path="example/prompt_audio.wav"# Change directory to the root directory
cd "$root_dir" || exitsource sparktts/utils/parse_options.sh# Run inference
python -m cli.inference \--text "${text}" \--device "${device}" \--save_dir "${save_dir}" \--model_dir "${model_dir}" \--prompt_text "${prompt_text}" \--prompt_speech_path "${prompt_speech_path}"

运行

bash infer.sh

能够成功合成语音，参数所表示的意思对应如下：

text 需要克隆语音所对应的文本
device 指定显卡编号
save_dir 克隆语音的保存路径
model_dir 大模型的保存路径
prompt_text 样本语音的文本信息
prompt_speech_path 样本语音的保存路径

python -m cli.inference \--text "你好" \--device 0 \--save_dir "result" \--model_dir pretrained_models/Spark-TTS-0.5B \--prompt_text "吃燕窝就选燕之屋，本节目由26年专注高品质燕窝的燕之屋冠名播出。豆奶牛奶换着喝，营养更均衡，本节目由豆本豆豆奶特约播出。" \--prompt_speech_path "example/prompt_audio.wav"

如果想要改变声音的性别，音调，速度，可以在infer.sh中看到对于这些参数的说明，但是这几个参数要一起加才能用，否则会报错