Featherweight Dictation expands to Windows
Featherweight Dictation has been available on Mac since earlier in 2026 and added Windows support in October. The product is built around a single design constraint: dictate in any app without the delay or resource overhead that typically comes with AI transcription.
The workflow is hold-to-dictate: hold a key, speak, release. Text appears at your cursor about half a second later. The model removes filler words, applies verbal corrections, formats numbers and symbols, and adapts the output style to the target app. A message in Slack gets different treatment than a code comment in Cursor or a draft in a writing tool.
On Mac, the app is 4.6 MB on disk, uses about 31 MB of memory, and 0% CPU when idle — measurements taken by the company on October 1, 2026. Speech is processed on Featherweight’s servers, not locally, so an internet connection is required.
Smart mode — which can rewrite selected text by voice, not only insert new text — is currently Mac-only. The Windows release covers the core hold-to-dictate workflow, with Smart mode on Windows described as forthcoming.
Pricing uses a free tier with a weekly word allowance. Paid plans remove the limit. The company says each computer gets one free account and no credit card is required to start.
For writers who move between many apps throughout the day, Featherweight’s low system footprint is a practical advantage over heavier tools that load voice models locally. The Windows release brings this to PC-based workflows.