<# 
Text-to-Speech WAV File Generator
Uses Windows Speech Platform API to convert text to speech and save as WAV

Author: Claude AI from prompts by Mike Shellim. With input from Paul Grey.

Requirements
============
Windows 10, 11
(optional) FFMpeg

To install this script
======================
1. Change the file extention of this script file to .ps1 (note numeral '1').
2. Create a folder to hold script file and output file 
3. Right click script file and choose 'Run with Powershell'
4. At the prompt, type in the text to be converted

Stripping silences (optional)
=============================
The script offers to strip leading and trailing silences, usesful for time critical setups.
This feature requires FFmpeg to be installed.
https://ffmpeg.org/download.html#build-windows
The 'essentials' pack is sufficient
  
Troubleshooting
===============
If the script flashes briefly then closes, it means the script has failed. Cause 
is almost certainly an overly restrictive PowerShell execution policy. This is easy to fix, 
see:

https://powershellfaqs.com/powershell-running-scripts-is-disabled-on-this-system/

Method 2 is recommended as a good balance between utility and security. 
#>

# Load required assemblies
Add-Type -AssemblyName System.Speech
Add-Type -AssemblyName System.Windows.Forms

function Get-InstalledVoices {
    $synthesizer = New-Object System.Speech.Synthesis.SpeechSynthesizer
    $voices = $synthesizer.GetInstalledVoices()
    $voiceList = @()
    
    foreach ($voice in $voices) {
        $voiceInfo = [PSCustomObject]@{
            Name = $voice.VoiceInfo.Name
            Gender = $voice.VoiceInfo.Gender
            Age = $voice.VoiceInfo.Age
            Culture = $voice.VoiceInfo.Culture.Name
            Enabled = $voice.Enabled
        }
        $voiceList += $voiceInfo
    }
    
    $synthesizer.Dispose()
    return $voiceList
}

function Show-VoiceMenu {
    param($voices)
    
    Write-Host "`nAvailable Voices:" -ForegroundColor Green
    Write-Host "=================" -ForegroundColor Green
    
    for ($i = 0; $i -lt $voices.Count; $i++) {
        $voice = $voices[$i]
        $status = if ($voice.Enabled) { "[Available]" } else { "[Disabled]" }
        Write-Host "$($i + 1). $($voice.Name) - $($voice.Gender), $($voice.Culture) $status"
    }
    Write-Host ""
}

function Generate-SpeechWav {
    param(
        [string]$text,
        [string]$voiceName,
        [string]$outputPath
    )
    
    try {
        # Create speech synthesizer
        $synthesizer = New-Object System.Speech.Synthesis.SpeechSynthesizer
        
        # Set the voice
        $synthesizer.SelectVoice($voiceName)
        
        # Set volume to maximum (100) for loudest output without distortion
        $synthesizer.Volume = 100
        
        # Set speaking rate to normal
        $synthesizer.Rate = 0
        
        # Configure WAV output with 16kHz sample rate
        $format = New-Object System.Speech.AudioFormat.SpeechAudioFormatInfo(16000, [System.Speech.AudioFormat.AudioBitsPerSample]::Sixteen, [System.Speech.AudioFormat.AudioChannel]::Mono)
        $synthesizer.SetOutputToWaveFile($outputPath, $format)
        
        # Speak the text (this generates the WAV file)
        $synthesizer.Speak($text)
        
        # Clean up
        $synthesizer.Dispose()
        
        return $true
    }
    catch {
        Write-Error "Error generating speech: $($_.Exception.Message)"
        if ($synthesizer) {
            $synthesizer.Dispose()
        }
        return $false
    }
}

function Test-FFmpegAvailable {
    try {
        $null = & ffmpeg -version 2>$null
        return $true
    }
    catch {
        return $false
    }
}

function Strip-AudioSilence {
    param(
        [string]$inputPath,
        [string]$outputPath
    )
    
    try {
        # FFmpeg command to strip leading and trailing silence
        # silencedetect: detect silence periods
        # silenceremove: remove silence from beginning and end

        $tempo = 1.0 # atempo value. Alters speed without altering pitch. Value in range 0.5 to 2.0
        $silence = 0.01

        $ffmpegArgs = @(
            "-i", $inputPath,
            "-af", "silenceremove=start_periods=1:start_silence=$silence :start_threshold=-50dB,areverse,silenceremove=start_periods=1:start_silence=$silence :start_threshold=-50dB,areverse,atempo=$tempo,volume=4dB",
            "-ar", "16000",
            "-ac", "1",
            "-y",
            $outputPath
        )
        
        Write-Host "Stripping silence using FFmpeg..." -ForegroundColor Yellow
        & ffmpeg @ffmpegArgs 2>$null
        
        if (Test-Path $outputPath) {
            return $true
        } else {
            return $false
        }
    }
    catch {
        Write-Error "Error stripping silence: $($_.Exception.Message)"
        return $false
    }
}

function Get-SafeFileName {
    param([string]$text)
    
    # Remove invalid filename characters and limit length to 40 characters
    $safeText = $text -replace '[^\w\s-]', '' -replace '\s+', '_'
    if ($safeText.Length -gt 40) {
        $safeText = $safeText.Substring(0, 40)
    }
    return $safeText
}

# Main script execution
Write-Host "==================================" -ForegroundColor Cyan
Write-Host "Text-to-Speech WAV File Generator" -ForegroundColor Cyan
Write-Host "==================================" -ForegroundColor Cyan
Write-Host ""

# Get available voices
Write-Host "Loading available voices..." -ForegroundColor Yellow
$voices = Get-InstalledVoices | Where-Object { $_.Enabled -eq $true }

if ($voices.Count -eq 0) {
    Write-Error "No enabled voices found on this system."
    exit 1
}

# Prompt for text input
do {
    $text = Read-Host "`nEnter the text to convert to speech"
    if ([string]::IsNullOrWhiteSpace($text)) {
        Write-Host "Text cannot be empty. Please try again." -ForegroundColor Red
    }
} while ([string]::IsNullOrWhiteSpace($text))

# Show voice menu and get selection
Show-VoiceMenu -voices $voices

do {
    $voiceSelection = Read-Host "Select a voice (1-$($voices.Count))"
    
    if ($voiceSelection -match '^\d+$' -and [int]$voiceSelection -ge 1 -and [int]$voiceSelection -le $voices.Count) {
        $selectedVoice = $voices[[int]$voiceSelection - 1]
        break
    } else {
        Write-Host "Invalid selection. Please enter a number between 1 and $($voices.Count)." -ForegroundColor Red
    }
} while ($true)

# Generate output filename
$safeFileName = Get-SafeFileName -text $text
$timestamp = Get-Date -Format "yyyyMMdd_HHmmss"
$outputFileName = "${safeFileName}_${timestamp}.wav"
$outputPath = Join-Path -Path (Get-Location) -ChildPath $outputFileName

# Display generation info
Write-Host "`nGeneration Details:" -ForegroundColor Green
Write-Host "==================" -ForegroundColor Green
Write-Host "Text: $text"
Write-Host "Voice: $($selectedVoice.Name)"
Write-Host "Output: $outputFileName"
Write-Host "Sample Rate: 16,000 Hz"
Write-Host "Format: 16-bit Mono WAV"
Write-Host ""

# Generate the WAV file
Write-Host "Generating speech..." -ForegroundColor Yellow

if (Generate-SpeechWav -text $text -voiceName $selectedVoice.Name -outputPath $outputPath) {
    Write-Host "Success! WAV file generated: $outputFileName" -ForegroundColor Green
    
    # Show file info
    if (Test-Path $outputPath) {
        $fileInfo = Get-Item $outputPath
        Write-Host "File size: $([math]::Round($fileInfo.Length / 1KB, 2)) KB" -ForegroundColor Cyan
        Write-Host "Full path: $($fileInfo.FullName)" -ForegroundColor Cyan
        
        # Prompt for silence stripping
        Write-Host ""
        $stripSilence = Read-Host "Would you like to strip leading and trailing silences? (y/N)"
        
        if ($stripSilence -match '^[Yy]') {
            # Check if FFmpeg is available
            if (Test-FFmpegAvailable) {
                # Create output filename with _short suffix using reliable method
                $baseName = [System.IO.Path]::GetFileNameWithoutExtension($outputFileName)
                $shortFileName = "${baseName}_short.wav"
                $shortOutputPath = Join-Path -Path (Get-Location) -ChildPath $shortFileName
                
                if (Strip-AudioSilence -inputPath $outputPath -outputPath $shortOutputPath) {
                    Write-Host "Success! Silence-stripped - file generated: $shortFileName" -ForegroundColor Green
                    
                    # Show comparison info
                    $shortFileInfo = Get-Item $shortOutputPath
                    Write-Host "Original file size: $([math]::Round($fileInfo.Length / 1KB, 2)) KB" -ForegroundColor Cyan
                    Write-Host "Stripped file size: $([math]::Round($shortFileInfo.Length / 1KB, 2)) KB" -ForegroundColor Cyan
                    $savings = [math]::Round((($fileInfo.Length - $shortFileInfo.Length) / $fileInfo.Length) * 100, 1)
                    Write-Host "Size reduction: $savings%" -ForegroundColor Cyan
                } else {
                    Write-Host "Failed to strip silence from audio file." -ForegroundColor Red
                }
            } else {
                Write-Host "FFmpeg is not available on this system." -ForegroundColor Red
                Write-Host "Please install FFmpeg to use the silence stripping feature." -ForegroundColor Yellow
                Write-Host "You can download FFmpeg from: https://ffmpeg.org/download.html" -ForegroundColor Yellow
            }
        }
        
        # Ask if user wants to play the file
        $play = Read-Host "`nWould you like to play the generated file? (y/N)"
        if ($play -match '^[Yy]') {
            try {
                Start-Process -FilePath $outputPath -Wait:$false
                Write-Host "Playing audio file..." -ForegroundColor Green
            }
            catch {
                Write-Host "Could not play the file automatically. You can play it manually: $outputFileName" -ForegroundColor Yellow
            }
        }
    }
} else {
    Write-Host "Failed to generate WAV file." -ForegroundColor Red
    exit 1
}

Write-Host "`nScript completed successfully!" -ForegroundColor Green