Since FFmpeg 0.10 release FFV1 codec supports alpha channel which is getting
enabled when using PIX_FMT_RGB32 pixel format. This leads to incompatibility of
videos rendered in Blender with almost all external players (especially in OSX).
Seems that PIX_FMT_BGR0 is recommended to be used to make videos compatible with
older players which doesn't support alpha channel in FFV1.
Also added an option to switch to RGBA rendering if FFV1 codec is used and if RGBA
rendering is used FFV1 will be using PIX_FMT_RGB32 format which supports alpha channel.
Most part of this commit fixes issues with FFmpeg output with
currently supported codecs:
- avcodec_encode_video might return zero which doesn't mean error
happened, but blender will handle this as error and will stop
rendering to video file.
- Changing output video codec wouldn't update "expert" options
set for video output which leads to some sideeffects like
ignored Lossless option for x264 codec.
This fixes allowed to add QTRLE codec easily.
Issue was caused by incorrectly set PTS value frames came form Scene strip renderer.
This value used to be calculated from RenderData current and start frame which
lead to non-uniformuly counting which totally confuses encoder.
Switch append_avi and append_ffmpeg to use current frame from rendering scene
(which was already passing to this functions and was used mostly for logging)
and start frame of rendering scene (it's new parameter added). This allowed to
calculate correct PTS value easily and get rid of global static sframe variable
in writeavi.c file.
Several issues were discovered when was looking into this bug:
- MPEG file format didn't set needed codec settings such as frame
rate and so, Was caused by not very correct fix fix#21351.
- "Expert" codec settings stored in idprops was affected on
formats which don't actually need them causing some conflicts
in codec settings.
- Flash codec doesn't support b-frames.
Now C presets shouldn't affect on each other and flash coded wouldn't
use b-frames even when using h264 format. Should work fine for files
created from scratch. If existing files fails to render, try to
switch file format to something else and then back to needed value.
- remove redundant casts
- replace strcmp's with "" to just check first char.
- added WM_event_print(), debug mode only to print events since the structs values are not that meaningful.
- added warnings if locale/font dirs cant be found.
Ogg format does support only vorbis, theora, speex and flac audio codecs.
Added check for result of av_write_header() and show info in header about
error while initializing streams.
This commit also fixed crash when using vorbis audio format.
It used to be floating point exception. SOlved by initializing
audio_stream->codec->time_base with proper rational value as it's
done in FFmpeg sources.
* replace by BLI_snprintf in various places, note _snprintf on windows
does not properly null terminate the string.
* fix overflow in sequencer proxy code due to buffer being smaller than
specified size.
* fix some usage of snprintf as strcpy, this is will go wrong if the
string contains % characters.
* remove BLI_dynstr_printf function in gpu module, use BLI_dynstr_appendf
2 changes
- When writing OGG only allow Theora encoding, this fixes the crash.
- When setting the MPEG preset, dont allow the 'Codec' to be left as Theora, this is just confusing.
* note that this is highly confusing for users and devs - there are 4 places to set the codec/format, with both python and C presets :S.
- Sequencer dynamics: Now it's possible to change the output channels and the resampling quality also increased (previously maximum quality was 44,1 kHz)
- Changed two buffers to use ffmpeg allocation, not sure if that helps somehow.
Added central compatibility header file, which enables blender to compile
against very old ffmpeg versions as well as very new versions using the
*NEW* API. (Old API functions are simulated using macros and inline functions)
Added a whole lot of additional checks, tested against 6 different versions
down the timeline, hopefully, now finally all is well.
Added some API compatibility code again, since some API-changes weren't even documented
(they even didn't do a proper version-bump, arghh!)
If it breaks again, please tell!
* removed a lot of old cruft code for ancient ffmpeg versions
* made it compile again against latest ffmpeg / libav GIT
(also shouldn't break distro ffmpegs, since those API changes
have been introduced over a year ago. If it nevertheless breaks,
please send me an email)
Both stored the filename of the blend file, but G.sce stored the last opened file.
This will make blender act differently in some cases since a relative path to the last opened file will no longer resolve (which is correct IMHO since that file isnt open and the path might not even be valid anymore).
Tested linking with durian files and rendering to relative paths when no files is loaded however we may need to have some operators give an error if they are used on the default startup.blend.
