--starttime option fails to seek AVCHD movies and produces garbage at beginning
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| ffmpeg2theora |
New
|
Undecided
|
Unassigned | ||
Bug Description
When encoding with option --starttime, risk is high that frame available in source movie at the targeted position won't be a key frame, but rather a predicted frame. In case source movie is an AVCHD 1080i stream (file extension 'm2ts' or 'mts'), OGV file displays a scrambled picture over a couple of frames, at the very beginning, when created with -s option. Likely, ffmep2theora starts working from AVCHD B-frames, not I-frame. A few frames later, Theora picture gets rebuilt rightly. May such a behaviour be a consequence of a lack of IDR I-frames in the AVCHD stream delivered by the camcorder (here a Canon) ? A screenshot taken from the flawed OGV file is attached to the bug report.
Test environment:
OS.X 10.5.8
ffmpeg2theora 0.27
Quicktime 7.6.6 + Perian 1.2.1
VLC 1.0.3
| Golffies (golffies) wrote | #1 |
| summary: |
- --starttime option produce garbage at beginning + --starttime option produces garbage at beginning |
| Golffies (golffies) wrote Re: --starttime option produces garbage at beginning | #2 |
| summary: |
- --starttime option produces garbage at beginning + --starttime option fails to seek AVCHD movies and produces garbage at + beginning |
| Golffies (golffies) wrote | #4 |
| Golffies (golffies) wrote | #5 |
I forgot to report that ffmpeg2theora fails to seek the AVCHD file to the position targeted by --starttime option, but starts encoding from the beginning of the AVCHD source file, delivering scrambled frames at the beginning of ogv file. Such a behaviour seems constant with AVCHD files used as input, whatever the value given to --starttime option might have been.