Sapi Tts

Windows SAPI5 text-to-speech with Neural voices. Lightweight alternative to GPU-heavy TTS - zero GPU usage, instant generation. Auto-detects best available voice for your language. Works on Windows 10/11.

Security Scan

The skill is an instruction-only Windows PowerShell script that uses the built-in SAPI5/.NET speech APIs to produce WAV files — its requirements and actions are coherent with its stated purpose.

Purpose & Capability

Name/description claim Windows SAPI5 TTS; the SKILL.md supplies a PowerShell script that uses System.Speech and PresentationCore to list voices, pick a voice, synthesize to WAV, and optionally play — these are exactly the capabilities needed for on‑device SAPI5 TTS. No unrelated credentials, binaries, or external services are requested.

Instruction Scope

Instructions stay within TTS scope (enumerate installed voices, select best match, write WAV, optionally play). The script writes output files under $env:USERPROFILE\.openclaw\workspace\tts (creates that directory if missing) — benign for a local TTS tool but worth noting. Also the SKILL.md and script target Windows 10/11, whereas registry metadata lists no OS restriction (minor metadata mismatch).

Install Mechanism

No install spec and no code files beyond SKILL.md; user is asked to save the provided PowerShell script — lowest-risk installation model (nothing is downloaded or executed automatically by the registry).

Credentials

No environment variables, credentials, or external config paths are required. The script references only $env:USERPROFILE to build a local output path, which is appropriate for generating files in the user's workspace.

Persistence & Privilege

Skill is not always-enabled, does not request elevated privileges, does not modify other skills or system configurations beyond creating its own output directory and files. Agent autonomous invocation remains default but is not combined with broad access here.

Assessment

This is an instruction-only Windows PowerShell TTS script that appears coherent and low risk. Before installing: (1) review the tts.ps1 contents (the SKILL.md contains the full script) to confirm you’re comfortable with it writing WAV files to $USERPROFILE\.openclaw\workspace\tts; (2) run it only on Windows 10/11 (it uses .NET System.Speech and PresentationCore); (3) if you prefer not to keep generated files, specify a custom -Output path or periodically clean the folder; (4) ensure your PowerShell execution policy/privileges are appropriate (you may need to unblock the script), and (5) if you have any security policy forbidding running third‑party scripts, treat this like any other user-provided script and run it in a controlled environment or sandbox first.

SAPI5 TTS (Windows)

Lightweight text-to-speech using Windows built-in SAPI5. Zero GPU, instant generation.

Installation

Save the script below as tts.ps1 in your skills folder:

<#
.SYNOPSIS
    Windows SAPI5 TTS - Lightweight text-to-speech
.DESCRIPTION
    Uses Windows built-in speech synthesis (SAPI5).
    Works with Neural voices (Win11) or legacy voices (Win10).
    Zero GPU usage, instant generation.
#>

param(
    [Parameter(Mandatory=$false, Position=0)]
    [string]$Text = "",
    
    [Parameter(Mandatory=$false)]
    [Alias("Voice", "v")]
    [string]$VoiceName = "",
    
    [Parameter(Mandatory=$false)]
    [Alias("Lang", "l")]
    [string]$Language = "fr",
    
    [Parameter(Mandatory=$false)]
    [Alias("o")]
    [string]$Output = "",
    
    [Parameter(Mandatory=$false)]
    [Alias("r")]
    [int]$Rate = 0,
    
    [Parameter(Mandatory=$false)]
    [Alias("p")]
    [switch]$Play,
    
    [Parameter(Mandatory=$false)]
    [switch]$ListVoices
)

Add-Type -AssemblyName System.Speech
$synth = New-Object System.Speech.Synthesis.SpeechSynthesizer

$installedVoices = $synth.GetInstalledVoices() | Where-Object { $_.Enabled } | ForEach-Object { $_.VoiceInfo }

if ($ListVoices) {
    Write-Host "`nInstalled SAPI5 voices:`n" -ForegroundColor Cyan
    foreach ($v in $installedVoices) {
        $type = if ($v.Name -match "Online|Neural") { "[Neural]" } else { "[Legacy]" }
        Write-Host "  $($v.Name)" -ForegroundColor White -NoNewline
        Write-Host " $type" -ForegroundColor DarkGray -NoNewline
        Write-Host " - $($v.Culture) $($v.Gender)" -ForegroundColor Gray
    }
    Write-Host ""
    $synth.Dispose()
    exit 0
}

if (-not $Text) {
    Write-Error "Text required. Use: .\tts.ps1 'Your text here'"
    Write-Host "Use -ListVoices to see available voices"
    $synth.Dispose()
    exit 1
}

function Select-BestVoice {
    param($voices, $preferredName, $lang)
    
    if ($preferredName) {
        $match = $voices | Where-Object { $_.Name -like "*$preferredName*" } | Select-Object -First 1
        if ($match) { return $match }
        Write-Warning "Voice '$preferredName' not found, auto-selecting..."
    }
    
    $cultureMap = @{
        "fr" = "fr-FR"; "french" = "fr-FR"
        "en" = "en-US"; "english" = "en-US"
        "de" = "de-DE"; "german" = "de-DE"
        "es" = "es-ES"; "spanish" = "es-ES"
        "it" = "it-IT"; "italian" = "it-IT"
    }
    $targetCulture = $cultureMap[$lang.ToLower()]
    if (-not $targetCulture) { $targetCulture = $lang }
    
    $neuralMatch = $voices | Where-Object { 
        $_.Name -match "Online|Neural" -and $_.Culture.Name -eq $targetCulture 
    } | Select-Object -First 1
    if ($neuralMatch) { return $neuralMatch }
    
    $langMatch = $voices | Where-Object { $_.Culture.Name -eq $targetCulture } | Select-Object -First 1
    if ($langMatch) { return $langMatch }
    
    $anyNeural = $voices | Where-Object { $_.Name -match "Online|Neural" } | Select-Object -First 1
    if ($anyNeural) { return $anyNeural }
    
    return $voices | Select-Object -First 1
}

$selectedVoice = Select-BestVoice -voices $installedVoices -preferredName $VoiceName -lang $Language

if (-not $selectedVoice) {
    Write-Error "No SAPI5 voices found! Install voices in Windows Settings > Time & Language > Speech"
    $synth.Dispose()
    exit 1
}

if (-not $Output) {
    $ttsDir = "$env:USERPROFILE\.openclaw\workspace\tts"
    if (-not (Test-Path $ttsDir)) { New-Item -ItemType Directory -Path $ttsDir -Force | Out-Null }
    $timestamp = Get-Date -Format "yyyyMMdd_HHmmss"
    $Output = "$ttsDir\sapi_$timestamp.wav"
}

try {
    $synth.SelectVoice($selectedVoice.Name)
    $synth.Rate = $Rate
    $synth.SetOutputToWaveFile($Output)
    $synth.Speak($Text)
    $synth.SetOutputToNull()
    
    Write-Host "Voice: $($selectedVoice.Name) [$($selectedVoice.Culture)]" -ForegroundColor Cyan
    Write-Host "MEDIA:$Output"
    
    # Auto-play if requested (uses .NET MediaPlayer, no external player)
    if ($Play) {
        Add-Type -AssemblyName PresentationCore
        $player = New-Object System.Windows.Media.MediaPlayer
        $player.Open([Uri]$Output)
        $player.Play()
        Start-Sleep -Milliseconds 500
        while ($player.Position -lt $player.NaturalDuration.TimeSpan) {
            Start-Sleep -Milliseconds 100
        }
        $player.Close()
    }
    
} catch {
    Write-Error "TTS failed: $($_.Exception.Message)"
    exit 1
} finally {
    $synth.Dispose()
}

Quick Start

# Generate audio file
.\tts.ps1 "Bonjour, comment vas-tu ?"

# Generate AND play immediately
.\tts.ps1 "Bonjour !" -Play

Parameters

Parameter	Alias	Default	Description
-Text	(positional)	required	Text to speak
-VoiceName	-Voice, -v	auto	Voice name (partial match OK)
-Language	-Lang, -l	fr	Language: fr, en, de, es, it...
-Output	-o	auto	Output WAV file path
-Rate	-r	0	Speed: -10 (slow) to +10 (fast)
-Play	-p	false	Play audio immediately after generation
-ListVoices			Show installed voices

Examples

# French with auto-play
.\tts.ps1 "Bonjour !" -Lang fr -Play

# English, faster
.\tts.ps1 "Hello there!" -Lang en -Rate 2 -Play

# Specific voice
.\tts.ps1 "Salut !" -Voice "Denise" -Play

# List available voices
.\tts.ps1 -ListVoices

Installing Neural Voices (Recommended)

Neural voices sound much better than legacy Desktop voices.

Windows 11

Neural voices are built-in. Go to: Settings → Time & Language → Speech → Manage voices

Windows 10/11 (More voices)

For additional Neural voices (like French Denise):

1. Install NaturalVoiceSAPIAdapter
2. Download voices in Settings → Time & Language → Speech
3. Run -ListVoices to verify

Performance

• Generation: Instant (< 1 second)
• GPU: None
• CPU: Minimal
• Quality: Good (Neural) / Basic (Legacy)

Credits

Made by Pocus 🎩 — AI assistant, with Olive (@Korddie).