Speech Handler

Windows desktop app for turning speech into text and text into speech. Transcribe audio files or live input with a local Vosk model, OpenAI Whisper, or ElevenLabs, then play or save the transcript with Kokoro or Piper neural voices.

Features

• Transcribe audio files to plain text and SubRip (.srt) subtitles.
• Live transcription from a microphone or system-audio (loopback) source, with an input-level meter.
• Choose a transcription engine
  • Vosk (offline): pick a language and model size, and download models from Transcription settings.
  • OpenAI Whisper: whisper-1, gpt-4o-mini-transcribe, or gpt-4o-transcribe, with optional translation to English.
  • ElevenLabs: scribe_v2 or scribe_v1.
• Vosk memory cache: set how much RAM to reserve for loaded models, keep several in memory so switching is instant, and unload models you no longer need.
• Spell checking on the transcript and subtitle editors, with suggestions, Ignore All, and Add to Dictionary. Custom words are stored per language.
• Save and copy the transcript as .txt or subtitles as .srt.
• Speak the transcript with downloaded Kokoro or Piper voices, including speaking-speed control.
• Save spoken audio of the transcript to a file.
• SSML tab for controlling how the transcript is spoken: pauses (<break>), rate and volume (<prosody>), approximate <emphasis>, <sub>, <p> and <s>. It starts as a copy of the transcript, and a tag list inserts the supported tags at the cursor. Speak and Save audio use the SSML while that tab is selected.

Requirements

• Windows 10 or 11, 64-bit
• .NET 10 SDK to build from source

Local Vosk models and Kokoro/Piper voice packs download from inside the app. OpenAI and ElevenLabs need an API key (entered in settings, or OPENAI_API_KEY / ELEVENLABS_API_KEY). Keys are not written to disk.

Models, voices, dictionaries, and settings live under %LOCALAPPDATA%\SpeechHandler.

Build and run

dotnet run --project SpeechHandler/SpeechHandler.csproj

A manual GitHub Actions workflow can publish a win-x64 zip from Actions → Build and Release.

License

Speech Handler is licensed under the PolyForm Noncommercial License 1.0.0. You may use, modify, and share it for noncommercial purposes. Commercial use requires a separate license from the copyright holder.