I can't stabilize any 360 footage using the 3D Camera Tracker from VR Comp Editor. 95% of the time, I run the analysis to create the camera for stabilization, it fails with the incredibly non-descriptive and undocumented error message "unable to acquire rendered frame." The error appears for videos of any length and any resolution. It also appears at various progress stages of the process and even within the same file (it can be at 5%, or at 55%, or at 98% completion).
At this point, I've exhausted any possible solution I could find:
- Reduced the Comp width down to 260 pixels and aspect ratio to 1:1: it speeds up processing but doesn't guarantee success even with a 6-second video
- Converted the video to sequences of JPEGs and PNGs: the success rate is slightly higher (the process is likely to complete than not) but this slows down the process exponentially (a 20-second video takes 5+ minutes to get converted into images then takes longer to analyze and re-encoded)
- Played with every combination of settings when creating the 3D Edit
- Disabled the antivirus (Windows Defender)
- Turned off all cloud syncing and took the computer offline
- Moved the source files to an M.2 SSD from RAID0 (2x7200rpm WD Black drives)
- Moved disk cache to the M.2 SSD
- Increased paging file size to 16GB
- Tried changing from CUDA to OpenCL and back
The source camera is Nikon KeyMission 360. The videos are stitched in-camera and exported as equirectangular 2:1. Tested video resolution was 3840×1920 and 3840×2160.
Computer specs: Dell XPS 8930, i7-8700, 32GB RAM @ 2667MHz, GeForce GTX 1070 8GB, AE 16.0.0 build 235 with 429 plugins.
Every single tutorial talks about how easy this feature is and has no tips for troubleshooting. I'm hoping an actual Adobe (or Mettle) developer chimes in on this.
