2.28.3

This week was mostly about housekeeping and some QOL changes.

NEW

• Negative sync offset: Perhaps the most requested feature is here. You can now set negative offset for a video, i.e. video starts before telemetry stream. The exported video shows no overlay for a part that has no data (if there is any). The editor always shows widgets, even if there are no data, with placeholder values - to avoid confusion why your widgets are disappearing.

IMPROVEMENTS

• Auto sync: Turns out video creation time in metadata tag can be both UTC or local time disguised as UTC. Thanks to @samarskyrider-dev for bringint this issue up. OVRLEY now treats this time as ambiguous, calculates potential syncing offsets for both scenarios. If only one options is valid, it will be used. If both are valid, it defaults to local time. Added a switch to syncing panel that allows user to toggle the timezone application to the video creation time.
• Manual sync: You can now select clip/activity in the timeline and manually adjust the sync offset by left/right arrow keys for high precision. Autoscroll speed during dragging the clip is now slower at higher timeline zoom levels, so you have better control over the clip placement.
• Video creation time: OVRLEY now prioritizes GPS timestamps from protobuf to determine video creation time if available; ffprobe is only used as fallback. This should make syncing more robust since the times come from satellites and should be more reliable.
• Export quality: Enabled variable bitrate where possible - this improves quality and reduce blockiness around the overlay text in highly dynamic scenes. I am considering going even further and use constant quality flag, but that would give only abstract quality values and unpredictable file sizes (since you can't use CRF with bitrate).
• Preview performance: Reduced re-renders, geometry and widget rebuilds, so playback and scrubbing should be smoother.
• Widget resizing: Consolidated the widget resizing logic, so it should be more consistent, less jittery and more performant.

BUG FIXES

• Fixed GPS timestamps and video creation time for DJI AC004 remote controller. Both are now correctly treated as naive/UTC, respectively. Also applied 8 hour offset since the GPS timestamps appear to encoded in local Chinese timezone as per data I got from Philip. This remains to be monitored - I have no idea if this weird offset is general behavior of DJI AC004 or unit dependent. (Does anyone know?)
• Fixed minor jagged edges of remaining route map line in the render - the geometry is now rotated prior to rasterization.
• Fixed improperly rotated video frames when encoding with Intel QSV. After extensive testing, it turned out that Vpp_qsv=transpose is no-op, so the overlay frame is rotated instead outside the filter.
• Fixed audio for trimmed clips not being interpreted correctly by some media players (looking at you, Windows Media Player) and resulting in malformed audio.
• SRT timestamps are properly treated as naive instead of UTC; giving correct timestamps/time widget values.
• Fixed datetime(utc) in AirData .CSV not parsing properly; the time widget now populates correctly.
• Fixed a bug where sync offset section would not display for AirData .CSV.
• Fixed interpolation bug where some metrics did not respect null/0 values and bridged the gap between two valid values instead of leaving it empty.