Roadmap

Roadmap & Known Gaps

This page lists the parts of TypeWhisper for Linux that are visible in the UI or described in older documentation but are not yet fully active in this branch. Treat anything here as planned work rather than a shipped feature.

Rule of thumb: any feature not described as active in the documentation should be treated as pending until it is implemented in this repository. The old README described broader platform coverage than this branch currently ships.

Known gaps

In-place self-update

Downloading and applying a new build from within the app is not implemented. TypeWhisper does check GitHub Releases for a newer version — the Check for Updates button on the About page, plus a once-per-day startup check, surface a non-obtrusive banner when an update is available — but installing the update is still a manual download from GitHub Releases. See Installation for the download formats.

Marketplace / store browsing

Marketplace/store browsing is intentionally not active in the Linux UI right now. Plugin update handling is limited compared with the intended full marketplace workflow. See Plugins.

Marketplace-style model browsing

Marketplace-style model browsing and management is not wired up in the Linux UI yet. Model selection comes from whichever bundled or manually installed plugins are present and enabled — see Transcription engines.

Windows release channels / Velopack

Windows release channels and Velopack update-channel controls are not used by this Linux branch.

Planned: a model tuned for dictation cleanup

The current Text cleanup at the Medium and High levels is a general-purpose LLM steered by a long system prompt. The maintainer plans to train — or fine-tune — a model purpose-built for dictation cleanup so the behavior lives in the model weights instead of being carried by the prompt. The goal is cleanup that is faster, more consistent, and far less sensitive to prompt wording than leaning on a general model.

Product direction

The direction for this branch is a dependable local voice-writing layer for Linux: dictation that reads like finished writing, reliable text insertion with clear fallback behavior, voice-native editing of selected text, app-aware style and formatting, strong Linux/developer workflows, and local-first data ownership. The project deliberately keeps its local-first, Linux-native identity rather than adopting cloud sync, team/enterprise dashboards, or screen-capture context.

Related pages

• Text cleanup — the cleanup levels available today
• Plugins — plugin and model management state
• About — the update-check banner and Error Log

---

Changelog

Date	Change
2026-06-17	Initial version.
2026-06-19	Removed "Interface language switching" — shipped in v0.9.0 (live-switching UI language). See General Settings.