In case we receive a flush event before having our caps set, we will
end up trying to create a theora encoder even though we are not ready.
Avoid that situation making sure we are initialized before accepting to
be flushed.
https://bugzilla.gnome.org/show_bug.cgi?id=709858
Instead, remember we need a keyframe, and we will force the encoder
to emit one next time we submit a new frame.
Since libtheora does not have an API to request a keyframe, we reset
the max keyframe interval to 1 temporarily.
This has the advantage that the rate control keeps its history,
and that the encoder won't choose different quant tables or
somesuch, thus requiring new streamheaders (although this is
probably only a theoretical possibility). Should also be a
bit faster than resetting the encoder.
https://bugzilla.gnome.org/show_bug.cgi?id=663350
No need to copy buffers we put into the streamheader any more
now that we don't put caps on buffers any more, so there's no
danger of a refcount cycle.
FIXME : Don't forget to backport changes that happened to theoraenc
since April 2011
theoraenc: Don't create keyframe on time gap
There is no rational to do so, and also gst_video_encoder_set_discont() is
gone from base class.
This option will produce duplicate frames if we get
a frame with GAP flag. This will reduce CPU load and file size.
This option should be disabled for real time applications, because it
collects GAP frames and waits until it gets a non GAP frame to start
encoding.
v30.06.2011: make some spell changes.
v03.07.2011: add handling of EOS and discontinuous for dup-on-gap.
v19.12.2011: fix pointer dangling in theora_timefifo_free
v20.12.2010: fix timestamp bug for dup-on-gap=0
Bugzilla: https://bugzilla.gnome.org/show_bug.cgi?id=627459
Signed-off-by: Oleksij Rempel (Alexey Fisher) <bug-track@fisher-privat.net>
If both quality and bitrate are set, libtheora will try to meet
both constraints, causing it to prefer emitting a smaller number
of good frames, to emitting the full number of frames that would
not meet the requested quality. This causes a slideshow effect
when the bitrate is low and the quality is high. And the default
theoraenc is high (48/63).
So only set quality when it is requested, and leave it unset
otherwise.
https://bugzilla.gnome.org/show_bug.cgi?id=658443
The speed-level property, which allows callers to trade of encoding
quality for speed in the libtheora api, has a version-dependent
maximum and default values. Instead of hardcoding the acceptable
range for the theoraenc element's presentation of this setting,
we query the library directly at class initialization time and
set the maximum and default values from that. If the query fails,
we fall back to the previous default setting.
To keep the values reported by gst-inspect (which I'm told use
the spec values from the class) with those available on an\
instantiated element, we remove to setting of enc->speed_level
from the initializer and instead pass G_PARAM_CONSTRUCT to
the property spec flags, asking g_object to set this property
when theoraenc objects are constructed.
NB in theory the maximum speed-level could depend on the actual
video caps. If later versions of libtheoraenc do this, a second
call will need to be made from theora_enc_reset to update the
property, since this function is mostly useful for realtime
adjustment of performance while the pipeline is running.
libtheora has two encoding modes, CBR, where it tries to hit a target
bitrate and VBR where it tries to achieve a target quality.
Internally if the target bitrate is set to anything other then 0 the
encoding-mode is CBR.
This means that the gstreamer element can leave the video_quality
setting alone as long as the user is tweaking the bitrate. Which has the
nice side-effect that if the user explicitely sets the bitrate to 0
(which is actually the default), the quality value doesn't get reset and
one ends up encoding VBR at quality-level 0...
Remove "This property requires libtheora version >= 1.1" qualifiers
from property descriptions. They aren't needed any longer now that
we require libtheora >= 1.1.