| Age | Commit message (Collapse) | Author |
|---|
|
there's a long delay at the end of the job while they are written.
We must still write FULL frames even if the queue is long (we only
stop doing that if the queue has too many FULL frames i.e. too much
memory consumption). With this commit we stop writing REPEAT/FAKE
frames when the queue gets long and assume there will always be a
sequence image for writing and hence the main writer thread will reduce
the queue given time.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
decoding in general).
|
|
|
|
|
|
Apparently we have to pass frame_rate as a parameter when making the buffer
source for the filter chain.
|
|
to put subtitles in.
|
|
|
|
|
|
rendering subtitles for preview or burn-in.
|
|
video when they change (#1290).
|
|
|
|
|
|
|
|
|
|
were merged.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
onto frame boundaries. ceil() can creep upwards with rounding error,
which I have seen happen over a save/load.
|
|
so that we don't have to run through the whole of the trimmed
portion to get to the data we want.
|
|
the resulting images; fixes #1274.
|
|
|
|
|
|
|
|
|
|
things.
|
|
|
|
|
|
|
|
|
