Skrivi Lytt 0.4.3 - Test release

Skrivi Lytt 0.4.3 — Test release

A clearer reading interface, consistent with Skrivi Snakk.

• A short first-run introduction explains reading selected text anywhere and reading a screen region.
• Current shortcuts are prominent; shortcut changes use a separate, cancellable capture dialog.
• Settings and the reader fit smaller windows; scrolling over controls no longer changes their values.
• A simpler main screen keeps audio saving and screen-region reading, without the sample button or model-routing detail.
• Consistent Privacy/About screens and simpler installer licence information.

Use Skrivi-Lytt-0.4.3-windows-x64-setup.exe for normal installation. Quit the existing tray reader first. The full installer includes the English and Bokmål voices, signed engines and OCR. Existing models, settings and saved audio are preserved. The runtime ZIP is a build dependency, not an installer. The MSIX is for Microsoft Store submission; availability here does not imply Store certification.

These are the exact signed files approved after PC testing, built from bc84d5f in https://github.com/workavoidance/Skrivi-STT/actions/runs/35537490895. Signatures, checksums, both speech engines and installer preservation checks passed.

Companion: Skrivi Snakk 0.3.2.

Skrivi Lytt 0.4.1 - Test release

Skrivi Lytt 0.4.1 — signed test release

Skrivi TTS is now Skrivi Lytt, alongside the dictation app Skrivi Snakk.
Existing model, runtime, settings and audio locations stay unchanged.

Download

• Skrivi-TTS-0.4.1-windows-x64-setup.exe: signed Windows installer for normal use.
• Skrivi-TTS-0.4.1-Store.msix: associated package for uploading to the reserved
  Skrivi Lytt listing in Microsoft Partner Center. This is not a sideload installer.
• Other ZIP/JSON/checksum files are developer build inputs and verification evidence.

Quit the existing reader from its tray menu before running setup. The full installer
includes both speech models and offline OCR. It preserves matching cached files,
settings, presets and saved audio, and refuses to overwrite conflicting files.

Reader improvements

• A rounded light/dark progress pill appears immediately after a reading hotkey.
• Separate capture, OCR, voice startup/loading, speech preparation and playback
  stages. Loading explains that the voice stays ready for subsequent reading.
• Escape cancels active work; the pill does not take focus from the source app.
• Two approved models: Piper Talesyntese for Bokmal and Kokoro for English.
• Optional automatic OCR columns; paragraph/single-column remains the default.
• Windows-default/English/Bokmal interface controls, opt-in sign-in startup and
  user-initiated update checks. No automatic update or model-download checks.
• Signed application, native speech/OCR components, installer and uninstaller.

Verification

Built source: 410df3850ec32176bc6578d9e0dd12f900d0925a.
Signed build and installation checks.

21 unit checks and light/dark/focus/cancellation UI checks passed. Clean-machine
installation, shortcut migration, conflict refusal, reinstall/uninstall preservation
and both voices passed. Local checks verified the installer signature, all 4,795
Store payload hashes, 4,280 unchanged runtime files, both voices cold/warm and OCR.

The Store package uses Skrivi.SkriviLytt and the reserved publisher identity.
Microsoft Store certification and installed Store-package testing remain pending.
Upload instructions and certification notes.

The earlier 0.4.0 UNASSOCIATED package must not be uploaded. This release retains
exactly the signed speech runtime bytes from 0.4.0; native voices/defaults are unchanged.

Skrivi TTS 0.4.0 — signed reading and OCR

Skrivi TTS 0.4.0 adds immediate feedback when reading selected text or a screen region: a light/dark activity pill shows capture, recognition, voice startup/loading, speech preparation and playback. Escape cancels active work.

The reader offers Talesyntese Bokmål and Kokoro English, an optional column-layout mode, Windows-default/English/Bokmål interface language, sign-in startup controls and a manual update check. No automatic update checks or speech/OCR network calls are added.

Download Skrivi-TTS-0.4.0-windows-x64-setup.exe. Quit the reader, then run setup. Use the full installer once when moving from the previous unsigned engines. Matching existing model files are reused; models and settings remain outside app versions. The separate signed-runtimes ZIP is a developer build input, not an installer.

The app, executable libraries, native helpers, installer and uninstaller are signed. Valid upstream signatures are preserved. CI verified 370 payload signatures, both signed speech engines, installed startup, conflicting-file protection, reinstall and uninstall preservation of 4,291 data files, and the installed uninstaller signature. A clean Windows build avoids the earlier Qt DLL conflict.

The Store validation package was built successfully but is UNASSOCIATED and not for upload. A separate Skrivi TTS listing has not yet been reserved. Its official identity must be applied before building the submission package; certification has not been run.

Build source: 6f27073fff6e83c12c08b2abec7b2e8a0105d451.
Successful signed build and Store validation.

This is a test release for real-world feedback. Source, build recipes, test evidence and continuity notes are maintained in this repository.

0.3.0 - Offline screen-region reading (test update)

Adds fully local screen-region reading for clean printed text. Press Ctrl+Alt+Shift+Space, drag around a paragraph or column on the pointer monitor, and release to read immediately. Escape cancels selection. Recognised text is available in the reader for review.

This is a test update for an existing full 0.2.1 installation. Extract the ZIP, run INSTALL.bat, then quit the old tray app and reopen. Includes OCR runtime and English/Norwegian OCR data; reuses existing speech models without downloading them. Version 0.2.1 remains available.

Validation: 22/24 synthetic development crops exact; 12/12 separate held-out clean/JPEG/dark crops exact. Roughly 0.2 seconds per packaged OCR request on the development PC, excluding speech. Installed startup and OCR passed; 25 pre-existing data/model/settings files preserved. Crop geometry, cancellation and stale-result checks passed. Source: 9d67200.

Select one paragraph or column; cross-monitor selections, complex page layout and HDR are not validated. Nynorsk OCR does not imply a Nynorsk voice. Binaries remain unsigned; this release does not resolve the separately reported Smart App Control issue. No security settings are changed.

Skrivi TTS 0.2.1 - local Norwegian and English reader

Windows setup executable

For the simplest first installation, download Skrivi-TTS-0.2.1-windows-x64-setup.exe, open it and follow the setup wizard. It includes the same app, runtimes and Norwegian/English voices as the full ZIP. Quit Skrivi TTS before installing. No administrator rights or internet connection are needed after download.

Setup offers Norwegian and English, an optional desktop shortcut, and an uninstaller that keeps models, reusable runtimes, settings and saved audio. Existing conflicting model/runtime files are preserved and cause setup to stop. The app remains unsigned; Windows may show a reputation warning. No Microsoft Store version is available yet.

Installer recipe: bf612d1. Clean Windows verification passed installation, startup, reinstall, both speech engines, conflict protection and data-preserving uninstall. See INSTALLER-SHA256SUMS.txt and INSTALLER-PROVENANCE.json for this EXE; SHA256SUMS.txt still covers the unchanged original ZIPs.

---

Skrivi TTS is a separate open-source Windows reader with its own tray icon.

Includes Piper Talesyntese Norwegian and Kokoro Heart English. Select text and press Ctrl+Alt+Space to read immediately; press again to stop. Automatic chooses one language for the complete passage, with manual override.

Alternative ZIP downloads

• Full first installation: Skrivi-TTS-0.2.1-Windows-x64.zip (640 MB), including two models and runtimes. Extract and run INSTALL.bat.
• Existing full 0.2.1 installations: Skrivi-TTS-0.2.1-App-Update-Windows-x64.zip (40 MB). No model weights or runtimes; prerequisite checks protect existing installations. This is the format for future code-only updates.
• SHA256SUMS.txt verifies both archives.

Models and settings are preserved across updates. No installer network calls or background update checks. Project & updates opens GitHub when clicked. An explicit Check for updates action is recorded for the next release; this build was not rebuilt just for that request.

Corrects the pre-release 0.2.0 Qt/ICU DLL error. The installed executable startup, automatic language routing, both speech engines, playback completion and Stop passed. Full install and app-only reinstall preserved all 24 existing model/settings/voice/preset files and modification times. Eleven tests and a clean GitHub Windows build/startup check passed.

Only Talesyntese and Kokoro weights are bundled, not akhbar/chatterbox-tts-norwegian or the other optional models. Kokoro weights are Apache-2.0; the Piper voice repository declares MIT and the Talesyntese dataset is CC0. See docs/RELEASE_LICENSE_REVIEW.md, THIRD_PARTY_NOTICES.md and docs/SOURCES.md for precise upstream sources and runtime licensing.

App built from 8d1f24a; app-only packaging from 7ec23ba. Exact source provenance is recorded in package.json. Windows x64, unsigned. Image OCR is future work. Some protected/elevated applications require pasting text; broader manual UI accessibility testing remains outstanding.

Skrivi TTS 0.1.0

First maintained Windows release of Skrivi TTS, continuing the existing Norwegian shootout and VoxCPM2 Reader work in a separate repository.

Download the Windows ZIP, extract it, and double-click INSTALL.bat. The application appears in the Start menu and on the desktop. No separate Python installation is needed.

• Persistent model library: application updates preserve downloaded weights, presets, reference voices and saved audio.
• Four working CPU models: public VoxCPM2 Q4, Chatterbox Multilingual Q4 ONNX, Piper NVCC and Piper Talesyntese.
• Model-specific inference controls, reference WAV support, named presets and listening ratings.
• Models stay loaded between readings; results distinguish cold/warm runs and record exact settings and timings.
• Complete native WAV playback and cancellation of the owned engine process tree.

The ZIP is approximately 104 MB and contains the application and reusable runtimes, not model weights. Import existing model files or download a chosen model once from the Model library tab. All four models were imported from existing caches on the development PC without redownloading them.

Validation: five library/configuration tests; real packaged-runtime cold/warm synthesis for all four models; native WAV validation and cancellation; real reinstall preserved all 19 model/registry files. Verified on Ryzen 7 7700 / 32 GB RAM. GPU inference is not implemented. The existing Vox adapter retains its 4,096-character limit. No claim of a production-ready accessibility reader or model-weight training is made.

See README.md and PROJECT_STATUS.md for settings, provenance, build instructions and next experiments. Checksums are attached.