Skald for press

Short version (≈ 60 words)

Skald is a native dictation tool for macOS and Windows; the Linux client is in closed beta. Hold a hotkey, speak, release — and context-aware, cleaned-up text lands wherever the cursor is: mail, Slack, IDE, address bar. Skald reads the active app, window title, and field role to match tone and structure, and respects the speaker's language. Audio stays in memory; servers are in the EU.

Long version (≈ 220 words)

Skald is a native dictation tool that turns spoken words into typed text in any input field — a reply in Apple Mail or Outlook, a Slack DM, a Notion note, a commit message in the IDE, or a query in the address bar. To use it, hold a global hotkey, speak one to three sentences, and release: cleaned-up text appears at the cursor.

Unlike generic speech-to-text utilities, Skald reads the surrounding context. App identity, window title, browser URL, and the role of the focused field travel with the raw transcript to the cleanup model, which shapes the output accordingly — short and direct for a DM, with greeting and sign-off for a client email, no structural rewrites in code mode. There are three modes: Dictate replaces typing; Command replaces a highlighted selection following a voice instruction ("shorter", "in English", "as a list"); Brain-Dump records longer thoughts and returns structured Markdown.

Skald runs on macOS 13+ and Windows 11 (x64 and ARM64) with feature parity; the native Linux client is in closed beta. On Mac and Windows, speech recognition runs on-device where the hardware supports it; the cleanup model only ever sees text — audio never leaves the device. On Linux, Skald uses cloud STT on EU servers. There is no telemetry, no analytics, and no training on user voices. The publisher is Tiamat UG (haftungsbeschränkt), based in Ahrensburg, Germany.