rvcやsvc用の学習データを作成するためのスクリプト。

Create-Dataset.ps1

# Requires -Version 7.1

[CmdletBinding()]
param (
    [Parameter(Mandatory)]
    [string]
    $SrcDir,

    [Parameter(Mandatory)]
    [string]
    $DstDir,

    [Parameter()]
    [int]
    $SamplingRate = 44100,

    [Parameter()]
    [ValidateSet("left", "rigth", "mix")]
    [string]
    $MonoChannel = "left",

    [Parameter()]
    [double]
    $HPF = 80,

    [Parameter()]
    [boolean]
    $NormalizeDC = $true,

    # https://ffmpeg.org/ffmpeg-filters.html#anlmdn
    [Parameter()]
    [ValidateRange(0.00001, 10000)]
    [double]
    $NoiseReductionLevel = 0.01,

    [Parameter()]
    [int]
    $LoudnormI = -14,

    [Parameter()]
    [int]
    $LoudnormTP = -2,

    [Parameter()]
    [int]
    $LoudnormLRA = 11,

    [Parameter()]
    [ValidateRange(0, 97)]
    [int]
    $SilenceThreshold = -40,

    [Parameter()]
    [double]
    $MinSilenceDuration = 0.2,

    [Parameter()]
    [double]
    $MinSplitDuration = 2.0,

    [Parameter()]
    [double]
    $SkipTotalDuration = 0.4,

    [Parameter()]
    [double]
    $ConcatDuration = 2,

    [Parameter()]
    [ValidateSet("pcm_s16le", "pcm_s24le", "pcm_s32le", "pcm_f32le", "pcm_f64le")]
    [string]
    $OutputCodec = "pcm_s16le",

    [Parameter()]
    [ValidateSet("pcm_s16le", "pcm_s24le", "pcm_s32le", "pcm_f32le", "pcm_f64le")]
    [string]
    $TempCodec = "pcm_f32le",

    [Parameter()]
    [ValidateSet("quiet", "panic", "fatal", "error", "warning", "info", "verbose", "debug", "trace")]
    [string]
    $FFmpegLogLevel = "error"
)

$ErrorActionPreference = 'Stop'

$SrcDir = Resolve-Path $SrcDir
$DstDir = (mkdir $DstDir -Force).FullName

if (Get-ChildItem $DstDir) {
    Write-Host "[warn] File exists in destination directory." -ForegroundColor Yellow
}

# 論理コア数
$ncpus = (Get-CimInstance win32_processor).NumberOfLogicalProcessors
# $ncpus = (Get-ComputerInfo).CsNumberOfLogicalProcessors

# 並列実行数
$parallelLimit = [Math]::Max(1, [int]($ncpus * 0.8))

# Step 1. モノラル化 & ハイパスフィルター & ノイズリダクション
$i = 0
$inputFiles = Get-ChildItem $SrcDir -File
$inputFiles | ForEach-Object -ThrottleLimit $parallelLimit -Parallel {
    $DstDir = $using:DstDir
    $SamplingRate = $using:SamplingRate
    $MonoChannel = $using:MonoChannel
    $HPF = $using:HPF
    $NormalizeDC = $using:NormalizeDC
    $NoiseReductionLevel = $using:NoiseReductionLevel
    $TempCodec = $using:TempCodec
    $FFmpegLogLevel = $using:FFmpegLogLevel

    $inputFile = $_

    $outputFilePath = Join-Path $DstDir "$($inputFile.BaseName).wav"

    $panFilter = if ($MonoChannel -eq 'left') {
        'pan=mono|FC=FC+FL'
    }
    elseif ($MonoChannel -eq 'rigth') {
        'pan=mono|FC=FC+FR'
    }
    else {
        'pan=mono|FC=FC+FL*0.5+FR*0.5'
    }

    ffmpeg `
        -v $FFmpegLogLevel `
        -i $inputFile.FullName `
        -af "$($panFilter),highpass=f=$($HPF):n=$([int]$NormalizeDC),anlmdn=s=$($NoiseReductionLevel)" `
        -ac 1 -ar $SamplingRate -acodec $TempCodec `
        $outputFilePath `
        -y 2>&1 | Write-Host

    return $_
} | ForEach-Object {
    $i++
    Write-Progress "Step 1. Mono & HPF & Noise Reduction" "$i / $($inputFiles.Count)" -PercentComplete ($i / $inputFiles.Count * 100)
}

# Step 2. ラウドネス正規化
$i = 0
$inputFiles = Get-ChildItem $DstDir -File
$inputFiles | ForEach-Object -ThrottleLimit $parallelLimit -Parallel {
    $DstDir = $using:DstDir
    $SamplingRate = $using:SamplingRate
    $I = $using:LoudnormI
    $TP = $using:LoudnormTP
    $LRA = $using:LoudnormLRA
    $FFmpegLogLevel = $using:FFmpegLogLevel
    $TempCodec = $using:TempCodec

    $inputFile = $_

    $loudnormFilter = ffmpeg `
        -i $inputFile.FullName `
        -af "loudnorm=I=$($I):TP=$($TP):LRA=$($LRA):print_format=json" `
        -f null - 2>&1 `
    | Out-String `
    | Select-String "\[Parsed_loudnorm.*?\]\s*(\{[\s\S]*?\})" `
    | ForEach-Object { $_.Matches.Groups[1].Value } `
    | ConvertFrom-Json `
    | ForEach-Object {
        "loudnorm=I=$($I):TP=$($TP):LRA=$($LRA)" `
            + ":measured_I=$($_.input_i):measured_TP=$($_.input_tp):measured_LRA=$($_.input_lra)" `
            + ":measured_thresh=$($_.input_thresh):offset=$($_.target_offset)"
    }

    $outputFilePath = Join-Path $DstDir "$($inputFile.BaseName).temp.wav"

    ffmpeg `
        -v $FFmpegLogLevel `
        -i $inputFile.FullName `
        -af $loudnormFilter `
        -ac 1 -ar $SamplingRate -acodec $TempCodec `
        $outputFilePath `
        -y 2>&1 | Write-Host

    $inputFile | Remove-Item
    Rename-Item $outputFilePath -NewName "$($inputFile.BaseName).wav"

    return $_
} | ForEach-Object {
    $i++
    Write-Progress "Step 2. Loudness Normalization" "$i / $($inputFiles.Count)" -PercentComplete ($i / $inputFiles.Count * 100)
}

# Step 3. 無音で分割
$i = 0
$inputFiles = Get-ChildItem $DstDir -File
$inputFiles | ForEach-Object -ThrottleLimit $parallelLimit -Parallel {
    $DstDir = $using:DstDir
    $SilenceThreshold = $using:SilenceThreshold
    $MinSilenceDuration = $using:MinSilenceDuration
    $MinSplitDuration = $using:MinSplitDuration
    $FFmpegLogLevel = $using:FFmpegLogLevel

    $inputFile = $_

    $audioInfo = ffprobe `
        -v error `
        -i $inputFile.FullName `
        -print_format json -show_entries format=duration 2>&1 `
    | ConvertFrom-Json

    $totalDuration = [double]$audioInfo.format.duration

    # 最小時間より短い場合は分割しない。
    if ($totalDuration -lt $MinSplitDuration) {
        return $_
    }

    # 無音が検出された時間
    [double[]]$silenceTimes = ffmpeg `
        -i $inputFile.FullName `
        -af "silencedetect=n=$($SilenceThreshold)dB:d=$($MinSilenceDuration)" `
        -f null - 2>&1 `
    | Select-String "silence_end" `
    | ForEach-Object {
        $array = $_ -split ' '
        $silence_end = [double]$array[4]
        $silence_duration = [double]$array[7]
        # 無音部分の真ん中の時間を返す。
        return $silence_end - ($silence_duration / 2)
    }

    [double[]]$segmentTimes = @()

    # 分割後の長さが最小時間以上であれば、$segmentTimesに追加する。
    foreach ($st in $silenceTimes) {
        if (($st - [double]$segmentTimes[-1]) -ge $MinSplitDuration `
                -and ($totalDuration - $st) -ge $MinSplitDuration) {
            $segmentTimes += $st
        }
    }

    if (-Not $segmentTimes.Count) {
        return $_
    }

    $outputFilePath = Join-Path $DstDir "$($inputFile.BaseName)_%04d$($inputFile.Extension)"

    ffmpeg `
        -v $FFmpegLogLevel `
        -i $inputFile.FullName `
        -f segment -segment_times ($segmentTimes -join ',') `
        -reset_timestamps 1 `
        -map 0:a `
        -c:a copy $outputFilePath `
        -y 2>&1 | Write-Host

    Remove-Item $inputFile.FullName

    return $_
} | ForEach-Object {
    $i++
    Write-Progress "Step 3. Split By Silence" "$i / $($inputFiles.Count)" -PercentComplete ($i / $inputFiles.Count * 100)
}

# Step 4. 無音を削除
$i = 0
$inputFiles = Get-ChildItem $DstDir -File
$inputFiles | ForEach-Object -ThrottleLimit $parallelLimit -Parallel {
    $DstDir = $using:DstDir
    $SamplingRate = $using:SamplingRate
    $SilenceThreshold = $using:SilenceThreshold
    $MinSilenceDuration = $using:MinSilenceDuration
    $FFmpegLogLevel = $using:FFmpegLogLevel
    $OutputCodec = $using:OutputCodec

    $inputFile = $_

    # 最初の無音を削除
    $removeStart = "silenceremove=start_periods=1:start_silence=0.02:start_threshold=$($SilenceThreshold)dB:detection=rms:window=0.01"

    $filters = @(
        # デジタル無音を全て削除
        "silenceremove=window=0:detection=peak:stop_mode=all:start_mode=all:stop_periods=-1:stop_threshold=0",
        # 先頭の無音を削除
        $removeStart,
        # 末尾の無音を削除（逆転 → 無音除去 → 再逆転）
        "areverse", $removeStart, "areverse",
        # 一定の音量かつ一定の長さの無音を全て削除
        "silenceremove=stop_periods=-1:stop_duration=$($MinSilenceDuration):stop_threshold=$($SilenceThreshold)dB:detection=rms:window=0.01"
    )

    $outputFilePath = Join-Path $DstDir "$($inputFile.BaseName).temp.wav"

    ffmpeg `
        -v $FFmpegLogLevel `
        -i $inputFile.FullName `
        -af ($filters -join ',') `
        -ac 1 -ar $SamplingRate -acodec $OutputCodec `
        $outputFilePath `
        -y 2>&1 | Write-Host

    $inputFile | Remove-Item
    Rename-Item $outputFilePath -NewName "$($inputFile.BaseName).wav"

    return $_
} | ForEach-Object {
    $i++
    Write-Progress "Step 4. Remove Silence" "$i / $($inputFiles.Count)" -PercentComplete ($i / $inputFiles.Count * 100)
}

# Step 5. オーディオの長さを確認
$audioInfoList = [System.Collections.Generic.List[psobject]]::new()
$i = 0
$inputFiles = Get-ChildItem $DstDir -File
$inputFiles | ForEach-Object -ThrottleLimit $parallelLimit -Parallel {
    $DstDir = $using:DstDir

    $inputFile = $_

    $audioInfo = ffprobe `
        -v error `
        -i $inputFile.FullName `
        -print_format json -show_entries format=duration 2>&1 `
    | ConvertFrom-Json

    return [PSCustomObject]@{
        FullName = $_.FullName
        BaseName = $_.BaseName
        Duration = [double]$audioInfo.format.duration
    }
} | ForEach-Object {
    $audioInfoList += $_

    $i++
    Write-Progress "Step 5. Check audio duration" "$i / $($inputFiles.Count)" -PercentComplete ($i / $inputFiles.Count * 100)
}
$audioInfoList = @($audioInfoList | Sort-Object FullName)
$removeTargets = $audioInfoList | Where-Object { $_.Duration -le $SkipTotalDuration }
$concatTargets = $audioInfoList | Where-Object { $_.Duration -gt $SkipTotalDuration -and $_.Duration -le $ConcatDuration }

# Step 6. かなり短いファイルを削除
$i = 0
$removeTargets | ForEach-Object {
    $i++
    Remove-Item $_.FullName
    Write-Progress "Step 6. Remove very short files" "$i / $($removeTargets.Count)" -PercentComplete ($i / $removeTargets.Count * 100)
}

# Step 7. 短いファイルを連結
try {
    # 無音ファイルの作成
    $silentFilePath = Join-Path $DstDir "silent.temp.wav"
    ffmpeg `
        -v $FFmpegLogLevel `
        -f lavfi `
        -i "anullsrc=cl=mono:r=$($SamplingRate)" `
        -t 0.02 `
        $silentFilePath `
        -y 2>&1 | Write-Host

    $filePaths = [System.Collections.Generic.List[string]]::new()
    $i = 0
    while ($true) {
        $filePaths.Clear()
        $totalDuration = 0.0
        for (; $totalDuration -lt $ConcatDuration -and $i -lt $concatTargets.Count; $i++) {
            $t = $concatTargets[$i]
            $filePaths.Add($t.FullName)
            $totalDuration += $t.Duration
        }

        if (-not $filePaths.Count) {
            break
        }

        $outputFilePath = Join-Path $DstDir "$(Split-Path $filePaths[0] -LeafBase)_$($i).concat.wav"

        $fileList = ($filePaths | ForEach-Object { "file '$($_)'" }) -join "`nfile '$($silentFilePath)'`n"
        $fileList | ffmpeg `
            -v $FFmpegLogLevel `
            -protocol_whitelist 'pipe,file' `
            -safe 0 `
            -f concat `
            -i pipe:0 `
            -c:a copy $outputFilePath `
            -y 2>&1 | Write-Host

        $filePaths | Remove-Item

        Write-Progress "Step 7. Concat short files" "$i / $($concatTargets.Count)" -PercentComplete ($i / $concatTargets.Count * 100)
    }
}
finally {
    # 作成した無音ファイルを削除
    Remove-Item $silentFilePath -ErrorAction SilentlyContinue
}

# Step 8. リネーム
$i = 0
$dstFiles = Get-ChildItem $DstDir -File
$dstFiles | Sort-Object Name | ForEach-Object {
    $i++
    $_ | Rename-Item -NewName "$($_.BaseName).temp$($_.Extension)"
    Write-Progress "Step 8a. Rename files" "$i / $($dstFiles.Count)" -PercentComplete ($i / $dstFiles.Count * 100)
}
$i = 0
$dstFiles = Get-ChildItem $DstDir -File
$padWidth = $dstFiles.Count.ToString().Length
$dstFiles | Sort-Object Name | ForEach-Object {
    $i++
    $_ | Rename-Item -NewName "$($i.ToString().PadLeft($padWidth, '0'))$($_.Extension)"
    Write-Progress "Step 8b. Rename files" "$i / $($dstFiles.Count)" -PercentComplete ($i / $dstFiles.Count * 100)
}