Montreal Forced Aligner Windows笔记

最前面

根据官方文档，mfa2.x在windows环境下，功能受限， 但是可以使用主要功能：对齐功能！

Conda 下载

https://mirrors.tuna.tsinghua.edu.cn/anaconda/miniconda/

Conda 配置文件

add_anaconda_token: True
add_pip_as_python_dependency: True
aggressive_update_packages:
  - ca-certificates
  - certifi
  - openssl
allow_non_channel_urls: False
allow_softlinks: False
always_copy: False
always_softlink: False
always_yes: None
anaconda_upload: None
auto_update_conda: True
changeps1: True
channel_alias: https://conda.anaconda.org
channel_priority: True
channels:
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
  - loopbio
client_ssl_cert: None
client_ssl_cert_key: None
clobber: False
create_default_packages: []
custom_channels:
  pkgs/main: https://repo.anaconda.com
  pkgs/r: https://repo.anaconda.com
  pkgs/pro: https://repo.anaconda.com
  pkgs/free: https://repo.anaconda.com
custom_multichannels:
  local: []
  defaults: ["https://repo.anaconda.com/pkgs/main", "https://repo.anaconda.com/pkgs/free", "https://repo.anaconda.com/pkgs/r", "https://repo.anaconda.com/pkgs/pro"]
default_channels:
  - https://repo.anaconda.com/pkgs/main
  - https://repo.anaconda.com/pkgs/free
  - https://repo.anaconda.com/pkgs/r
  - https://repo.anaconda.com/pkgs/pro
disallowed_packages: []
download_only: False
envs_dirs:
  - /home/mosheng/miniconda3/envs
  - /home/mosheng/.conda/envs
force: False
json: False
local_repodata_ttl: 1
max_shlvl: 2
migrated_channel_aliases: []
no_dependencies: False
non_admin_enabled: True
notify_outdated_conda: True
offline: False
override_channels_enabled: True
path_conflict: clobber
pinned_packages: []
pkgs_dirs:
  - /home/mosheng/miniconda3/pkgs
  - /home/mosheng/.conda/pkgs
proxy_servers: {}
quiet: False
remote_connect_timeout_secs: 40.0
remote_max_retries: 3
remote_read_timeout_secs: 100.0
report_errors: None
rollback_enabled: True
safety_checks: warn
shortcuts: True
show_channel_urls: True
ssl_verify: True
track_features: []
use_index_cache: False
use_pip: True
verbosity: 0
whitelist_channels: []

准备创建MFA环境

使用Conda的命令行工具

添加conda-forge频道

conda config --add channels conda-forge

安装某版本的MFA

默认版本

conda create -n aligner -c conda-forge montreal-forced-aligner

特定版本

版本号链接

conda create -n aligner_vb11 -c conda-forge montreal-forced-aligner=2.0.0b11

指定路径

如下所示， 新环境的名字叫做mfaligner

conda create -c conda-forge -p E:\Dev\miniconda3\envs\mfaligner  montreal-forced-aligner=2.0.0b11

conda激活/进入mfa环境

退出当前环境

conda deactivate

进入刚刚创建的名叫aligner的环境

conda activate aligner

设置环境变量

这个文件夹将会用来存储下载好的模型
手动下载模型必须设置此路径
以方便mfa利用模型

MFA模型资料下载

https://mfa-models.readthedocs.io/en/latest/

注意： 推荐手动下载，不用命令

下载好的模型

MFARootDir是设置的环境变量路径，请参考上面 设置环境变量

准备语料

lab文件就是wav文件的语音识别结果，只是文本文件.txt改了后缀名，改为.lab

MFA验证下载模型和语料的匹配性

mfa validate E:\Dev\corpus\speaker1 english_us_arpa english_us_arpa

对齐

mfa align corpus_path english english target_path

必看！重要的Github issue

OSError: sndfile library not found

https://github.com/MontrealCorpusTools/Montreal-Forced-Aligner/issues/426