Loki Studio v1.1.2

← All updates

NEW FEATURES

• Finalize Videos tab: New dedicated tab for mixing multiple audio tracks before YouTube upload. YouTube only plays audio track 1, so this feature combines all tracks (up to 4) with individual volume controls (-20dB to +20dB) into a single stereo track. Creates _final variant of videos. Hidden by default - enable in Power User Settings under "Advanced Features" at the top of the page.
• Timeline Editor "Merge Audio Tracks" option: New checkbox below "Render Final Video" button. When enabled, merges all audio tracks into a single track using the volume and enabled settings from the Audio Tracks section. Output uses _final suffix instead of _edited. Great for quick finalization without switching to the dedicated Finalize Videos tab.
• Track enable/disable for audio mixing: Both Finalize Videos tab and Timeline Editor now support excluding individual tracks from the final mix via checkbox. Disabled tracks show strikethrough label and "OFF" in the volume display.
• Instructor personality: Academic style that provides accurate summaries of course content without step-by-step breakdowns or excessive encouragement. Stays close to transcript, describes topics covered and learning outcomes in professional academic language.
• Filipino/Tagalog language support: Added Filipino (Tagalog) to transcription and translation languages. Now supports 36 translation languages via NLLB-200.
• Per-video settings persistence (_settings.json): * Saves YouTube video ID after successful upload * Stores all custom settings per video (VAD, language, personality, perspective) * Enables future "Update Only" mode for already-uploaded videos * Allows metadata updates without re-uploading the video file * Preserves upload history (channel, playlist, privacy, scheduled date) * Tracks processing timestamps (transcribed, metadata generated, thumbnail created)

IMPROVEMENTS

• Pipeline status now detects uploaded videos via _settings.json
• Dashboard can now show which videos have already been uploaded to YouTube
• Renamed "Validate" button to "Connect" in Upload tab for clearer user experience
• Simplified YouTube connection messages (e.g., "Channel connected" instead of technical credential terminology)

BUG FIXES

• Sliding Window and Full Segment captions now correctly use the Max Width setting. Caption box now properly fills the configured percentage (e.g., 80%) of video width. One Word Pop style retains dynamic sizing to fit individual words.
• YouTube channel connection now properly refreshes expired logins before connecting. Previously, opening Upload tab with an expired login would show an error. Now the login is refreshed automatically when you click Connect.
• Upload tab buttons (Remove, Connect) no longer get disabled when connection fails. Users can now retry or remove the channel after an error.
• EDL Import now correctly handles cross-dissolve transitions from Premiere Pro. Previously, dissolves between clips would assign the cut to the wrong clip. Now uses TO CLIP NAME for dissolves (the incoming clip).
• EDL Import now detects end-of-clip cuts for multi-clip series. Previously, only gaps at the beginning of clips were detected. Now compares the EDL's Source OUT time against actual source file durations to find footage cut from the end of clips.
• Batch transcription no longer crashes when video files are deleted mid-batch. Now checks file existence before processing each video. If a file is missing, attempts to find the original (strips _edited/_final suffixes). If original is also missing, logs an error and skips to the next video.
• Application Settings no longer reset to defaults when switching profiles. Previously, settings like NLLB model size and "CC On By Default" would revert to defaults. Fixed initialization order to prevent QML default values from overwriting saved settings during profile loads.
• Starting a new transcription batch after cancelling no longer crashes. Previously, if you cancelled a batch (especially with translation enabled) and started a new one, the app could crash due to the previous worker thread still running. Now properly waits for complete thread shutdown before starting.
• Upload tab now properly auto-checks videos when they become ready. Previously, if a video was unchecked because it wasn't ready (missing metadata or thumbnail), it would stay unchecked even after becoming ready. Now only preserves unchecked state for videos the user explicitly unchecked.

TECHNICAL DETAILS

### Video Settings System

• New VideoSettings struct in VideoDataIO.h with sub-structures: * VideoUploadInfo: YouTube video ID, URL, channel, playlist, privacy, schedule * VideoProcessingInfo: Timestamps for transcription, metadata, thumbnail * VideoTranscribeSettings: Whisper model, VAD settings, detected language * VideoMetadataSettings: Personality, perspective, LLM provider, translations
• Settings saved to <videoname>_settings.json in video directory
• Automatic atomic file writes (temp file + rename) for data safety
• PipelineStatus enhanced with hasSettings, isUploaded, youtubeVideoId fields
• UploadTab saves settings after successful YouTube upload

KNOWN ISSUES

(Document any known issues)

UPGRADE NOTES

• No breaking changes from v1.1.1

← All updates  ·  Source-of-truth changelog on GitHub