OpenTalkIt

An open-source reimplementation of the Microsoft Talk It! frontend, built with Avalonia UI and .NET 10.

Talk It! was a Windows speech synthesis application from the 1990s powered by the SoftVoice engine (TIBASE32.DLL). OpenTalkIt recreates its UI and behaviour as a modern cross-architecture desktop app, wrapping the original 32-bit DLL through the companion TiSpeech library.

Features

• 20 voice personalities with per-personality presets (pitch, speed, pitch quality, vocal effort) matching the original application
• Pitch quality control: Natural / Monotone / Sung
• Vocal effort control: Normal / Breathy / Whispered
• Language switching: English / Spanish (German supported by the engine)
• Pitch and speed adjustable per personality
• Async speech with Stop support
• WAV export via WASAPI loopback capture

Requirements

Requirement	Detail
OS	Windows only
Architecture	x86 — TIBASE32.DLL is a 32-bit native DLL; the app is compiled as x86
.NET	.NET 10
Native DLLs	TIBASE32.DLL, TIENG32.DLL, optionally TISPAN32.DLL

The native DLLs are not included. You need an original Talk It! installation.

Getting Started

1. Clone the repo
2. Copy TIBASE32.DLL, TIENG32.DLL (and optionally TISPAN32.DLL) into OpenTalkIt/DLLs/
3. Build and run:

dotnet run --project OpenTalkIt/OpenTalkIt.csproj

Project Structure

The solution spans three sibling repositories:

TiSpeech/              Core P/Invoke bindings and engine types
TiSpeech.Client/       Named-pipe client that talks to TiSpeech.Host
TiSpeech.Host/         32-bit out-of-process host that loads TIBASE32.DLL
OpenTalkIt/            Avalonia UI application (this repo)
  Models/
    PersonalityButtonModel.cs   Per-personality button state
    PersonalityPreset.cs        Preset record (pitch, speed, pitch quality, vocal effort)
    PersonalityPresets.cs       Static lookup table of OG Talk It! presets per personality
    Records.cs                  TalkParameters — snapshot of all settings passed to the engine
  ViewModels/
    MainWindowViewModel.cs      Composition root; wires preset application on personality change
    PersonalityControlViewModel Personality grid + pitch/speed controls
    ParameterControlViewModel   Pitch quality, vocal effort, language
    TalkControlViewModel        Text input, Talk/Stop/Export commands
  Views/Controls/
    PersonalityControl.axaml    4×5 personality grid with pitch and speed inputs
    ParameterControl.axaml      Radio button groups for pitch quality, vocal effort, language
    TalkControl.axaml           Text box and Talk It! / Stop / Export buttons
  Services/
    WavRecorder.cs              WASAPI loopback capture for WAV export
    Settings.cs                 JSON settings persistence (%APPDATA%/OpenTalkIt/)

TiSpeech Library

TiSpeech is a standalone .NET 10 library that wraps TIBASE32.DLL via P/Invoke. Because TIBASE32.DLL is 32-bit, the engine runs in a dedicated TiSpeech.Host process (x86 self-contained executable). The main app communicates with it through TiSpeech.Client over a named pipe. This lets the 64-bit Avalonia host and the 32-bit engine coexist without forcing the whole app to run as x86.

TiSpeech.Host handles engine lifecycle, speech synthesis, voice parameter setting, and async completion notification through a hidden Win32 message-only window (no Windows Forms dependency).

See TiSpeech/TiSpeech.md for the full API reference.

Key enums

Enum	Controls
TiPersonality	Voice character (0–19)
TiVoicingMode	Vocal effort: Normal / Breathy / Whispered
TiF0Style	Pitch contour: Natural / Monotone / Sung / Whispered-style
TiSpeakingMode	Token interpretation: Natural / Word / Spell / Number
TiLanguage	Runtime language switch

WAV Export

Since SoftVoice exposes no file-export API, WAV export works by capturing the system audio output (WASAPI loopback) while speech is playing. The captured audio is written to a WAV file at a location chosen via a save-file dialog. The last-used export folder is persisted in %APPDATA%/OpenTalkIt/settings.json.

Reverse Engineering Notes

The preset values and engine parameter mappings were confirmed through a combination of Ghidra static analysis of TIBASE32.DLL and dynamic analysis (API Monitor + x32dbg breakpoints) of the original Talk It! application.

Notable findings:

• Vocal effort (Normal / Breathy / Whispered) is controlled by SVSetVoicingMode (values 0–2), not SVSetGlottalSource as the internal string table might suggest
• Pitch quality (Natural / Monotone / Sung) maps to SVSetF0Style values 0 / 2 / 4
• SVSetGlottalSource is present in the DLL but is not called by the original UI's Vocal Effort buttons

License

This project contains no original Talk It! code or assets. The native DLLs (TIBASE32.DLL etc.) remain the property of their respective owners and are not redistributed here.