Releases: vaclavik-xyz/diktato-releases
Releases · vaclavik-xyz/diktato-releases
Release list
Diktāto v3.3.0
What's New
- Text now appears instantly. Dictated text is inserted as whole runs instead of one keystroke at a time, so it lands at once rather than typing itself out. Apps that struggled with the old approach, Telegram in particular, keep up.
- Direct insertion is the default. New installs insert at the cursor without touching your clipboard. Existing preferences are kept.
- Decide when the pill is on screen. A single choice replaces the old on/off switch: keep it resting on screen, show it only while dictating, or hide it entirely.
- Control the live transcript. Turn the running transcript on or off, and choose how many lines stay visible. The pill grows line by line as you speak instead of jumping to full height.
- Recover a lost pill. A new button restores the pill to its default position after it has been dragged off-screen or onto a disconnected display.
- Reach history from the pill. History moved out of the overflow menu into the pill's quick actions.
Improvements
- Dictating no longer makes the app stutter. The recording timer, live transcript, and audio level each update in isolation instead of re-rendering the main window ten times a second.
- Interface text resolves through cached bundles, and theme colours are no longer rebuilt on every colour lookup.
- Every pill state now shares one shape and outline, so failures and popovers no longer look like a different app.
- Popovers open above the pill, or flip below it when the pill sits near the top of the screen, instead of covering it.
- Pill settings apply immediately rather than after the next dictation.
- Clicks next to the pill reach the app underneath instead of landing on an invisible overlay.
Fixes
- English interface text no longer falls back to Czech for strings whose translation matches the source.
- The live transcript no longer clips its top line.
Full Changelog
Diktāto v3.2.0
What's New
- Record meetings your way. Choose an in-person, online, or hybrid meeting and let Diktāto capture the right microphone and system-audio sources for each situation.
- Keep every meeting organized. The new Meetings section brings searchable transcripts, editable titles, folders, notes, summaries, decisions, and action items into one dedicated workspace.
- Know who said what. Speaker-aware transcripts let you rename and merge speakers, link recurring participants, and keep corrected identities for future meetings.
- Take meeting results anywhere. Export a consistent transcript and meeting output directly from the meeting workspace.
- See local transcription while you speak. Cloud transcription can now show a responsive on-device preview when a compatible local Apple or Parakeet engine is available.
Improvements
- Online and hybrid meetings preserve separate microphone and system-audio tracks before producing a source-aware merged transcript.
- Long-running capture now writes audio incrementally to disk, recovers interrupted sessions more safely, and avoids UI stalls or dropped admitted buffers.
- System-audio recording now captures the correct process-tap stream and requests the required macOS permission explicitly.
- Parakeet transcripts preserve segment timestamps and overlapping timing bounds for more reliable speaker alignment.
- Onboarding now recognizes already downloaded local models and clearly distinguishes installed models from selectable downloads.
- The custom dictionary uses a neutral technical example instead of an unrelated brand name.
Full Changelog
Diktāto v3.1.0
What's New
- A clearer place to start. The new Home screen brings dictation, meeting recording, and media transcription together without turning History into a launcher.
- Understand what Diktāto changed. The new Processing area explains cleanup, Vibe, dictionary terms, and word replacements, while History shows evidence of the processing applied to each transcript.
- Track your progress. Statistics now include usage trends, estimated time saved, goals, streaks, achievements, and counts for Vibe and word replacements.
- Start recording with confidence. Choose from five subtle start sounds, preview them in Settings, or keep recording completely silent.
Improvements
- Transcription setup can now be postponed during onboarding, with clear guidance and guarded actions until a provider is configured.
- Local model recommendations now adapt to Apple Silicon and Intel hardware, while Diktāto Cloud account setup is separated from transcription and cleanup choices.
- AI cleanup settings more clearly explain the active model, connected services, and the limitation of Vibe with cloud transcription.
- Media transcription, dashboard cards, navigation, onboarding controls, keyboard shortcuts, and Czech localization are more compact and consistent.
- Dictionary terms and word replacements now explain their distinct roles and expose their effect after processing.
Full Changelog
Diktāto v3.0.4
What's New
- See what takes time. History now shows the transcription and cleanup models, providers, and processing times for each dictation.
- Follow your words live. The floating pill can display the live transcript while you speak and while Diktāto finishes processing it.
- Choose where text goes. Keep the app focused when dictation starts, or follow whichever app is active when dictation ends.
Improvements
- Refined the notch and floating indicator for a cleaner, more consistent recording experience.
- Primary buttons and onboarding choices now respect the selected Diktāto color theme.
- Fixed target-app preservation when the floating indicator is disabled.
- Replaced the confusing onboarding completion icon with the Diktāto app icon.
Full Changelog
Diktāto v3.0.3
What's New
- A cleaner Diktāto installer. The installation window is now more compact, with clearer spacing between the app, recording indicator, and Applications folder.
Improvements
- The installer uses a simplified recording waveform design without redundant branding.
- Finder icons stay correctly aligned around the centered recording indicator.
- The mounted installer volume now includes the app version in its name, making multiple releases easy to distinguish.
Full Changelog
Diktāto v3.0.2
What's New
- A clearer Diktāto installer. The DMG now uses a clean branded layout with a recording waveform that naturally guides the app into Applications.
Improvements
- The installer background is embedded portably and now appears reliably when the DMG is opened on another Mac.
- Release publishing now verifies the public DMG and automatically updates the matching Homebrew cask.
Full Changelog
Diktāto v3.0.1
What's New
- Clear local model preparation. Diktāto now shows an active loader and elapsed time while Whisper or Parakeet is being prepared for the first time instead of appearing stuck at a fixed percentage.
Improvements
- The onboarding AI enhancement screen remains fully visible after local model setup.
- Onboarding window sizing now prevents collapsed content while still adapting to compact displays.
- Czech and English preparation guidance explains that first-time model optimization can take several minutes and is cached for later use.
- The DMG installer now uses a branded drag-to-Applications layout with clear visual guidance.
Full Changelog
Diktāto v3.0.0
What's New
- Meet Diktāto. The app now uses its new name and identity everywhere, from the macOS bundle and settings to downloads and updates.
- See what the app is doing. Local model loading is visible throughout onboarding and Settings, including switching and recovery states.
- Understand required permissions. Permission status, guidance, and recovery actions are clearer in onboarding and Settings.
Improvements
- Modifier-only hotkeys recover more reliably after macOS permission changes, and status bar shortcuts stay synchronized with the active configuration.
- Onboarding controls follow native macOS sizing, accessibility, keyboard navigation, and theme behavior more closely.
- Optional privacy-friendly telemetry requires explicit consent, contains no dictated text or stable identifier, and stops immediately after opt-out.
- Main-window, sidebar, floating-indicator, cancellation, and model-loading state transitions are more reliable.
Full Changelog
Diktátor v2.1.2
What's New
- Revisit past dictations. Optionally retain successful dictation audio on this Mac and create a new transcript with another available speech model.
- Try a different cleanup. Reprocess the original transcript with another configured AI provider, model, style, or processing depth before choosing whether to use the result.
- Configure AI enhancement during onboarding. Connect a supported cloud or local provider from the first-run flow, with Qwen recommended for local Czech cleanup.
Improvements
- Diktátor now behaves like a regular Dock app by default while keeping its menu bar controls.
- Audio retention is off by default, stays local, and offers explicit 24-hour, 7-day, 30-day, or transcript-lifetime policies.
- History reprocessing never downloads a speech model silently and clearly explains when retained audio or a configured model is unavailable.
- Provider credentials and configurations remain stored when switching between hosted, local, and custom services, without requiring an app restart.
- Model availability, AI reasoning modes, Czech localization, transcription errors, and onboarding guidance are clearer.
- The floating dictation indicator, voice-responsive orange line, toast presentation, cancellation handling, and cleanup performance are more reliable.
Full Changelog
Diktátor v2.1.1
What's New
- Edit selected text by voice. Select text in any app, dictate the change you want, and let Diktator replace it safely.
- Use different styles for different apps. Assign enhancement instructions to individual apps and switch automatically as you work.
- Recover failed dictations. If transcription or output fails, Diktator preserves the latest recording so you can retry it instead of dictating again.
Improvements
- The redesigned dictation pill is cleaner, keeps the orange voice response, and renders its glow without clipped edges.
- Onboarding and settings are simpler, with clearer model and AI enhancement choices, including custom compatible providers.
- Streaming capture is more reliable at the beginning of a dictation and across rapid press, release, cancel, and restart sequences.
- Navigation, accessibility labels, and empty-state guidance are clearer while advanced controls stay out of the primary flow.