@jdobbs
Full backup with 1-pass encoding using CRF prediction and "Quicker encode for Extras".
For some short clips (marked with an asteriks) I get a very long CRF prediction sequence, like:
Code:
.......
- [22:16:01] Extracting A/V streams [VID_01758]
- [22:16:17] Reencoding video [VID_01758]
- Source Video: MPEG-4 (AVC), 1920x1080
- Rate/Length: 23.976fps, 1'678 frames
- [22:16:17] Performing CRF Prediction...
- Analyzing 14.65 7.83 4.41 2.71 1.85 1.43 1.21 1.11 1.05 1.03 1.01 1.01 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 [1.00]
- [22:19:21] Encoding using constant rate factor.
- Performing size-correcting second pass...
- [22:22:19] Video Encode complete
.......
At the end, the size-correcting second pass produces a reasonably sized result. Just wondering if this is normal, or is there possibly a bug?
Edit:
My conclusion:
The problematic clips are language specific intro clips which consisted predominantly of black backgrounds with few thin white letters (e.g. name of an actor).
The CRF prediction tried to boost the file size to the targeted size, which -- based on the 2% sample frames -- drove the CRF down to 1.00xxx
The size-correction (great!) kicked in because the encode was slightly oversized (the full clip had some "normal" pictures at the end).
Some of these clips were even encoded at crf=1.00xxx without the need for the size-correction kicking in
=> After all, the final result was great and filled the BD25 to 96%, which is just perfect
Could the CRF prediction perhaps be stopped once it gets below say 5.0? Anyway, I think this disc is an exception (Bridge of Spies)