Frequently asked questions.

InkSpoke is a cross-platform desktop Voice OS that converts speech to formatted, context-aware text and injects it directly into any active application. It works offline-first with optional cloud LLM support for enhanced refinement.

InkSpoke runs on macOS (Apple Silicon, Ventura 13 or later), Windows 10 1903+ (x64), and Linux (x64, as an AppImage). On-device transcription is GPU-accelerated with Metal on Apple Silicon and CUDA on Windows with an NVIDIA card; Linux currently runs it on CPU, so prefer a smaller or cloud model there.

Yes. InkSpoke is offline-first by design. Speech recognition uses local Whisper models and text refinement can run entirely on-device via LlamaSharp. No internet connection is required for core functionality.

The free tier includes speech-to-text, refinement, five workspaces, custom vocabulary, cross-device sync with optional end-to-end encryption, meeting recording, and the Whisper Small model on-device. Pro adds voice commands, unlimited workspaces, the larger on-device models, on-device refinement, bring-your-own-key providers, and priority support.

BYOK lets you point InkSpoke at your own provider account so usage is billed to you at cost, with no markup. The desktop app speaks the OpenAI chat-completions format, so any OpenAI-compatible endpoint works — OpenAI, Groq, Azure OpenAI, or a self-hosted server such as vLLM, Ollama or LM Studio via their OpenAI-compatible endpoints. BYOK requires Pro.

Pick on-device models for speech and refinement and your audio and text never leave the machine — dictation keeps working offline. Pick a cloud model and that request goes to the provider. Dictation history is stored locally and never uploaded. Two things do reach the network at every launch regardless: a device registration and an update check, neither of which can currently be turned off in-app; block the app at your firewall if you need silence on the wire. There is no analytics or crash-reporting telemetry at all.

InkSpoke supports 99+ languages for speech recognition via Whisper. LLM refinement quality varies by language but works well with all major languages. Custom vocabulary support helps with domain-specific terms.

No. Put your cursor in any text field, press the hotkey, speak, and press it again — the refined text is typed straight into the application you were already in. InkSpoke picks the delivery method per app, using simulated typing for terminals, editors and remote sessions and a clipboard paste elsewhere. If it does borrow the clipboard, your previous contents are put back afterwards.

Yes. InkSpoke works with any application that accepts text input, including VS Code, JetBrains IDEs, Xcode, and terminal emulators. Workspace-aware context helps produce code-appropriate output.

Download InkSpoke for your platform and install it. A short setup wizard covers microphone and accessibility permissions. Then put your cursor in any text field and press Alt+Space (Option+Space on macOS) to start dictating, and press it again to finish. Ctrl+Shift+Space works too, as a second binding for apps that already use Alt+Space.