Frequently asked questions

Yes. Download Invoke and use every feature free for 7 days. No credit card, no signup. After 7 days, enter a license key to keep using it.

Yes. CPU mode is included and works out of the box—it's just slower (a few seconds instead of sub-second). A CUDA-capable GPU makes transcription near-instant.

All 99+ languages that Whisper supports. English is the default, but you can configure any language or enable auto-detection.

Your audio always stays local. The reformatter calls an LLM API to clean up the transcript—you choose which provider (OpenRouter, Claude, etc.). For fully local operation, use Ollama or LM Studio, or turn on privacy mode to lock everything local with one click. No code files are ever uploaded.

Running Whisper gives you a transcript. Invoke wraps that in a workflow — push-to-talk hotkey, auto-paste at your cursor, LLM reformatting that reads your project context, prompt modes for Claude Code / Cursor / Windsurf / Copilot / Codex, custom vocabulary, and dictation history. Whisper is the engine; Invoke is the car.

For the Windows installer: Windows 10/11 (64-bit), 4 GB RAM, and a microphone. An NVIDIA GPU with CUDA 12.x is recommended but not required — CPU mode works out of the box. Internet needed for license activation and AI reformatter (unless you use Ollama or LM Studio). See full requirements.

Press a hotkey when you have an image on your clipboard (e.g., after Win+Shift+S). Invoke saves it as a PNG and pastes the file path into your terminal. This lets you share screenshots with Claude Code and other AI tools that accept image paths — no manual saving or typing file paths.

Invoke runs on Windows 10/11 (64-bit). An NVIDIA GPU is recommended for fastest transcription but not required — CPU mode works on any machine.