Release notes

Re-exports get dramatically faster: unchanged scenes are reused instantly instead of being re-encoded — fix one scene and re-export in seconds.

• Scene-level render cache: each scene's rendered video is cached, keyed on everything that affects its output (media, narration audio, subtitles, styles, resolution, motion). Unchanged scenes are reused instantly with a 'reusing previous render ⚡' progress message.
• A 'fix one scene out of thirty and re-export' cycle drops from a full re-encode to a single scene + concat — ideal for iterative review workflows and multilingual variants. Cache capped at 2 GB with oldest-first eviction, auto-invalidated on app updates.
• Media duration probes memoized per file, eliminating repeated analysis during export and scene switching. Scene and overlay lists rebuild in a single UI update for smoother editing on large decks.

SCORM 1.2 export ships straight into your LMS, plus scene-level recording, dual picture-in-picture overlays, and audio-first output.

• SCORM 1.2 package export for Moodle / Canvas / TalentLMS / Cornerstone and other major LMS. Auto-tracks completion at 95% viewed, resume-position seek, and session time — verified with zero parser warnings on Rustici SCORM Cloud and on Moodle 4.5.
• Per-scene in-app microphone recording (NAudio WaveIn). Switch among three narration sources (auto-voice TTS / recording / import) per scene with a single flag; the source in use is shown during export and preview.
• Bulk Narrator Assignment: edit language and narrator voice per slide in a single view — with jump-to-current, voice and modified-row filters, JSON import / export, and AI prompt copy.
• Dual picture-in-picture overlays plus a sign-language slot for accessibility-ready output, and a new 'Manual creation' ribbon group consolidating screen recording and step capture for tutorial videos.
• Refreshed scene editing panel: five audio sources unified into one row with Segoe MDL2 icons, two-line scene titles, and a wrapping action row so buttons are never cut off.
• Fixes: wipe-position slider now reflected in output and clamped on-screen, corrected negative recording-length display, and resolved progress-rollback and scene-preview status issues.

License stability fix, global rollout, and UI consistency improvements.

• Removed WMI CPU ID dependency from DeviceFingerprint, eliminating 'in use on another device' and 'license auto-deleted' incidents.
• UI-language-based LemonSqueezy store routing (Japanese → JPY JP store, others → USD global store).
• 30s HttpClient timeout to prevent freezes on slow international connections.
• Help ribbon reordered, Backstage labels unified, GPT Image 2.0 premium installer banners, locale-specific subtitle preview samples.

Two-way sync between subtitles and PowerPoint speaker notes. Two-person review/production workflow now possible.

• Subtitle ↔ PowerPoint speaker note bi-directional sync. Reviewers proofread subtitles inside PowerPoint while video producers tune narration in Insight Training Studio.
• Stabilized single-PPTX workflow (customXml now well-formed) and unified multi-language license dialog across INMV / INST via InsightCommon.
• Ribbon cleanup (Open split into PPTX/PDF, Recent dropdown), unified taskbar name, refreshed Welcome screen hero icon.

47 languages supported + 9-country worker visa language presets built in.

• 47 languages for narration and subtitles (VoiceVox / Microsoft Neural / ElevenLabs / Edge TTS).
• Built-in language presets for 9 specified-skill worker origin countries (Vietnam, Indonesia, Myanmar, Philippines, Thailand, Cambodia, Nepal, Mongolia, Laos).

Release build enabled + semantic translation + Language label unified.

• Semantic translation: contextual rather than literal word-for-word output.
• Release build now treats warnings as errors; non-impactful analyzer warnings moved to NoWarn.
• 'Toggle Language' renamed to 'Language' across the Insight Slide / Sheet / Doc / Cast / Agent suite for consistency.

Major UI/UX overhaul + 5-engine consolidation + PDF translation quality improvements.

• Both settings and target panels toggle via edge strips (VS Code style), with cost-estimate comparison across engines and Shift+/ shortcut overview.
• Translation engines consolidated to 5 (DeepL / Google / Azure / Claude / Gemini); NMT vs LLM use cases now explicit in the UI.
• PDF translation quality: dedicated TOC clustering, 50% color propagation threshold, iterative font-shrink to prevent text-box overflow.
• Fixed double-launch crash during translation and resolved Windows hang / 15 GB memory spike on PDF save.
• Added EULA: install-time consent page clarifying responsibility for BYOK, AI translation, PDF translation, and third-party APIs.

Zero-setup Python runtime with Office automation, RPA and unattended scheduled execution.

• Bundled Python runtime and core libraries (openpyxl / python-pptx / python-docx / pypdf / pyautogui / Pillow) — no install, no pip, no PATH.
• Run AI-written Python as-is; save scripts, organize into projects, and run multi-file jobs.
• Unattended scheduled execution (daily / weekly) for file cleanup, reporting and screen-level RPA.
• Built-in learning center with lessons in Japanese and English.