1 2017-01-09 v1.6.1 "Long Tailed Duck"
2 This release improves upon the VP9 encoder and speeds up the encoding and
6 This release is ABI compatible with 1.6.0.
9 Faster VP9 encoding and decoding.
10 High bit depth builds now provide similar speed for 8 bit encode and decode
11 for x86 targets. Other platforms and higher bit depth improvements are in
15 A variety of fuzzing issues.
17 2016-07-20 v1.6.0 "Khaki Campbell Duck"
18 This release improves upon the VP9 encoder and speeds up the encoding and
22 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
23 in vpx_image and some minor changes to the VP8_COMP structure.
25 The default key frame interval for VP9 has changed from 128 to 9999.
28 A core focus has been performance for low end Intel processors. SSSE3
29 instructions such as 'pshufb' have been avoided and instructions have been
30 reordered to better accommodate the more constrained pipelines.
32 As a result, devices based on Celeron processors have seen substantial
33 decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
34 decoding speed improved between 10 and 30%. Between Javan Whistling Duck
35 and Khaki Campbell Duck, it improved another 10 to 15%.
37 While Celeron benefited most, Core-i5 also improved 5% and 10% between the
40 Realtime performance for WebRTC for both speed and quality has received a
44 A number of fuzzing issues, found variously by Mozilla, Chromium and others,
45 have been fixed and we strongly recommend updating.
47 2015-11-09 v1.5.0 "Javan Whistling Duck"
48 This release improves upon the VP9 encoder and speeds up the encoding and
52 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
53 controls and adds a variety of VP9 controls for testing.
55 The vpxenc utility now prefers VP9 by default.
58 Faster VP9 encoding and decoding
59 Smaller library size by combining functions used by VP8 and VP9
62 A variety of fuzzing issues
64 2015-04-03 v1.4.0 "Indian Runner Duck"
65 This release includes significant improvements to the VP9 codec.
68 This release is ABI incompatible with 1.3.0. It drops the compatibility
69 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
73 Faster VP9 encoding and decoding
74 Multithreaded VP9 decoding (tile and frame-based)
75 Multithreaded VP9 encoding - on by default
76 YUV 4:2:2 and 4:4:4 support in VP9
77 10 and 12bit support in VP9
78 64bit ARM support by replacing ARM assembly with intrinsics
81 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
85 Frame Parallel decoding fails for segmented and non-420 files.
87 2013-11-15 v1.3.0 "Forest"
88 This release introduces the VP9 codec in a backward-compatible way.
89 All existing users of VP8 can continue to use the library without
90 modification. However, some VP8 options do not map to VP9 in the same manner.
92 The VP9 encoder in this release is not feature complete. Users interested in
93 the encoder are advised to use the git master branch and discuss issues on
97 This release is ABI and API compatible with Duclair (v1.0.0). Users
98 of older releases should refer to the Upgrading notes in this document
102 Get rid of bashisms in the main build scripts
103 Added usage info on command line options
104 Add lossless compression mode
106 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
107 Add option to disable documentation
108 configure: add --enable-external-build support
109 make: support V=1 as short form of verbose=yes
110 configure: support mingw-w64
111 configure: support hardfloat armv7 CHOSTS
112 configure: add support for android x86
113 Add estimated completion time to vpxenc
114 Don't exit on decode errors in vpxenc
115 vpxenc: support scaling prior to encoding
116 vpxdec: support scaling output
117 vpxenc: improve progress indicators with --skip
118 msvs: Don't link to winmm.lib
119 Add a new script for producing vcxproj files
120 Produce Visual Studio 10 and 11 project files
121 Produce Windows Phone project files
122 msvs-build: use msbuild for vs >= 2005
123 configure: default configure log to config.log
124 Add encoding option --static-thresh
127 Miscellaneous speed optimizations for VP8 and VP9.
130 In general, quality is consistent with the Eider release.
133 This release represents approximately a year of engineering effort,
134 and contains multiple bug fixes. Please refer to git history for details.
138 This release acts as a checkpoint for a large amount of internal refactoring
139 and testing. It also contains a number of small bugfixes, so all users are
140 encouraged to upgrade.
143 This release is ABI and API compatible with Duclair (v1.0.0). Users
144 of older releases should refer to the Upgrading notes in this
145 document for that release.
148 VP8 optimizations for MIPS dspr2
149 vpxenc: add -quiet option
152 Encoder and decoder speed is consistent with the Eider release.
155 In general, quality is consistent with the Eider release.
157 Minor tweaks to ARNR filtering
158 Minor improvements to real time encoding with multiple temporal layers
161 Fixes multithreaded encoder race condition in loopfilter
162 Fixes multi-resolution threaded encoding
163 Fix potential encoder dead-lock after picture resize
166 2012-05-09 v1.1.0 "Eider"
167 This introduces a number of enhancements, mostly focused on real-time
168 encoding. In addition, it fixes a decoder bug (first introduced in
169 Duclair) so all users of that release are encouraged to upgrade.
172 This release is ABI and API compatible with Duclair (v1.0.0). Users
173 of older releases should refer to the Upgrading notes in this
174 document for that release.
176 This release introduces a new temporal denoiser, controlled by the
177 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
178 currently take a strength parameter, so the control is effectively
179 a boolean - zero (off) or non-zero (on). For compatibility with
180 existing applications, the values accepted are the same as those
181 for the spatial denoiser (0-6). The temporal denoiser is enabled
182 by default, and the older spatial denoiser may be restored by
183 configuring with --disable-temporal-denoising. The temporal denoiser
184 is more computationally intensive than the spatial one.
186 This release removes support for a legacy, decode only API that was
187 supported, but deprecated, at the initial release of libvpx
188 (v0.9.0). This is not expected to have any impact. If you are
189 impacted, you can apply a reversion to commit 2bf8fb58 locally.
190 Please update to the latest libvpx API if you are affected.
193 Adds a motion compensated temporal denoiser to the encoder, which
194 gives higher quality than the older spatial denoiser. (See above
195 for notes on upgrading).
197 In addition, support for new compilers and platforms were added,
199 improved support for XCode
200 Android x86 NDK build
204 Changing resolution with vpx_codec_enc_config_set() is now
205 supported. Previously, reinitializing the codec was required to
206 change the input resolution.
208 The vpxenc application has initial support for producing multiple
209 encodes from the same input in one call. Resizing is not yet
210 supported, but varying other codec parameters is. Use -- to
211 delineate output streams. Options persist from one stream to the
214 Also, the vpxenc application will now use a keyframe interval of
215 5 seconds by default. Use the --kf-max-dist option to override.
218 Decoder performance improved 2.5% versus Duclair. Encoder speed is
219 consistent with Duclair for most material. Two pass encoding of
220 slideshow-like material will see significant improvements.
222 Large realtime encoding speed gains at a small quality expense are
223 possible by configuring the on-the-fly bitpacking experiment with
224 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
225 faster (ARM) depending on the number of threads and bitrate
226 settings. This technique sees constant gain over the 5-16 speed
227 range. For VC style input the loss seen is up to 0.2dB. See commit
228 52cf4dca for further details.
231 On the whole, quality is consistent with the Duclair release. Some
234 Reduced blockiness in easy sections by applying a penalty to
237 Improved quality of static sections (like slideshows) with
240 Improved keyframe sizing with multiple temporal layers
243 Corrected alt-ref contribution to frame rate for visible updates
244 to the alt-ref buffer. This affected applications making manual
245 usage of the frame reference flags, or temporal layers.
247 Additional constraints were added to disable multi-frame quality
248 enhancement (MFQE) in sections of the frame where there is motion.
251 Fixed corruption issues when vpx_codec_enc_config_set() was called
252 with spatial resampling enabled.
254 Fixed a decoder error introduced in Duclair where the segmentation
255 map was not being reinitialized on keyframes (#378)
258 2012-01-27 v1.0.0 "Duclair"
259 Our fourth named release, focused on performance and features related to
260 real-time encoding. It also fixes a decoder crash bug introduced in
261 v0.9.7, so all users of that release are encouraged to upgrade.
264 This release is ABI incompatible with prior releases of libvpx, so the
265 "major" version number has been bumped to 1. You must recompile your
266 applications against the latest version of the libvpx headers. The
267 API remains compatible, and this should not require code changes in most
271 This release introduces several substantial new features to the encoder,
272 of particular interest to real time streaming applications.
274 Temporal scalability allows the encoder to produce a stream that can
275 be decimated to different frame rates, with independent rate targetting
278 Multiframe quality enhancement postprocessing can make visual quality
279 more consistent in the presence of frames that are substantially
280 different quality than the surrounding frames, as in the temporal
281 scalability case and in some forced keyframe scenarios.
283 Multiple-resolution encoding support allows the encoding of the
284 same content at different resolutions faster than encoding them
288 Optimization targets for this release included the decoder and the real-
289 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
290 this release. Encoder improvements followed a curve where speeds 1-3
291 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
292 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
296 Encoder quality in the single stream case is consistent with the Cayuga
300 This release fixes an OOB read decoder crash bug present in v0.9.7
301 related to the clamping of motion vectors in SPLITMV blocks. This
302 behavior could be triggered by corrupt input or by starting
303 decoding from a P-frame.
306 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
307 This is an incremental bugfix release against Cayuga. All users of that
308 release are strongly encouraged to upgrade.
310 - Fix potential OOB reads (cdae03a)
312 An unbounded out of bounds read was discovered when the
313 decoder was requested to perform error concealment (new in
314 Cayuga) given a frame with corrupt partition sizes.
316 A bounded out of bounds read was discovered affecting all
317 versions of libvpx. Given an multipartition input frame that
318 is truncated between the mode/mv partition and the first
319 residiual paritition (in the block of partition offsets), up
320 to 3 extra bytes could have been read from the source buffer.
321 The code will not take any action regardless of the contents
322 of these undefined bytes, as the truncated buffer is detected
323 immediately following the read based on the calculated
324 starting position of the coefficient partition.
326 - Fix potential error concealment crash when the very first frame
327 is missing or corrupt (a609be5)
329 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
331 - Revert 1-pass CBR rate control changes (e961317)
332 Further testing showed this change produced undesirable visual
333 artifacts, rolling back for now.
336 2011-08-02 v0.9.7 "Cayuga"
337 Our third named release, focused on a faster, higher quality, encoder.
340 This release is backwards compatible with Aylesbury (v0.9.5) and
341 Bali (v0.9.6). Users of older releases should refer to the Upgrading
342 notes in this document for that release.
345 Stereo 3D format support for vpxenc
346 Runtime detection of available processor cores.
347 Allow specifying --end-usage by enum name
348 vpxdec: test for frame corruption
349 vpxenc: add quantizer histogram display
350 vpxenc: add rate histogram display
351 Set VPX_FRAME_IS_DROPPABLE
352 update configure for ios sdk 4.3
353 Avoid text relocations in ARM vp8 decoder
354 Generate a vpx.pc file for pkg-config.
355 New ways of passing encoded data between encoder and decoder.
358 This release includes across-the-board speed improvements to the
359 encoder. On x86, these measure at approximately 11.5% in Best mode,
360 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
361 On ARM Cortex A9 with Neon extensions, real-time encoding of video
362 telephony content is 35% faster than Bali on single core and 48%
363 faster on multi-core. On the NVidia Tegra2 platform, real time
364 encoding is 40% faster than Bali.
366 Decoder speed was not a priority for this release, but improved
367 approximately 8.4% on x86.
369 Reduce motion vector search on alt-ref frame.
370 Encoder loopfilter running in its own thread
371 Reworked loopfilter to precalculate more parameters
372 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
373 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
374 Removed redundant checks
375 Reduced structure sizes
376 utilize preload in ARMv6 MC/LPF/Copy routines
377 ARM optimized quantization, dfct, variance, subtract
378 Increase chrow row alignment to 16 bytes.
379 disable trellis optimization for first pass
380 Write SSSE3 sub-pixel filter function
381 Improve SSE2 half-pixel filter funtions
382 Add vp8_sub_pixel_variance16x8_ssse3 function
383 Reduce unnecessary distortion computation
384 Use diamond search to replace full search
385 Preload reference area in sub-pixel motion search (real-time mode)
388 This release focused primarily on one-pass use cases, including
389 video conferencing. Low latency data rate control was significantly
390 improved, improving streamability over bandwidth constrained links.
391 Added support for error concealment, allowing frames to maintain
392 visual quality in the presence of substantial packet loss.
394 Add rc_max_intra_bitrate_pct control
395 Limit size of initial keyframe in one-pass.
396 Improve framerate adaptation
397 Improved 1-pass CBR rate control
398 Improved KF insertion after fades to still.
399 Improved key frame detection.
400 Improved activity masking (lower PSNR impact for same SSIM boost)
401 Improved interaction between GF and ARFs
402 Adding error-concealment to the decoder.
403 Adding support for independent partitions
404 Adjusted rate-distortion constants
408 Removed firstpass motion map
409 Fix parallel make install
410 Fix multithreaded encoding for 1 MB wide frame
411 Fixed iwalsh_neon build problems with RVDS4.1
412 Fix semaphore emulation, spin-wait intrinsics on Windows
413 Fix build with xcode4 and simplify GLOBAL.
414 Mark ARM asm objects as allowing a non-executable stack.
415 Fix vpxenc encoding incorrect webm file header on big endian
418 2011-03-07 v0.9.6 "Bali"
419 Our second named release, focused on a faster, higher quality, encoder.
422 This release is backwards compatible with Aylesbury (v0.9.5). Users
423 of older releases should refer to the Upgrading notes in this
424 document for that release.
427 vpxenc --psnr shows a summary when encode completes
428 --tune=ssim option to enable activity masking
429 improved postproc visualizations for development
430 updated support for Apple iOS to SDK 4.2
431 query decoder to determine which reference frames were updated
432 implemented error tracking in the decoder
433 fix pipe support on windows
436 Primary focus was on good quality mode, speed 0. Average improvement
437 on x86 about 40%, up to 100% on user-generated content at that speed.
438 Best quality mode speed improved 35%, and realtime speed 10-20%. This
439 release also saw significant improvement in realtime encoding speed
442 Improved encoder threading
443 Dont pick encoder filter level when loopfilter is disabled.
444 Avoid double copying of key frames into alt and golden buffer
446 x86 sse2 temporal filter
447 SSSE3 version of fast quantizer
448 vp8_rd_pick_best_mbsegmentation code restructure
449 Adjusted breakout RD for SPLITMV
450 Changed segmentation check order
451 Improved rd_pick_intra4x4block
452 Adds armv6 optimized variance calculation
453 ARMv6 optimized sad16x16
454 ARMv6 optimized half pixel variance calculations
455 Full search SAD function optimization in SSE4.1
456 Improve MV prediction accuracy to achieve performance gain
457 Improve MV prediction in vp8_pick_inter_mode() for speed>3
460 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
461 also includes support for "activity masking," which greatly improves
462 SSIM at the expense of PSNR. For now, this feature is available with
463 the --tune=ssim option. Further experimentation in this area
464 is ongoing. This release also introduces a new rate control mode
465 called "CQ," which changes the allocation of bits within a clip to
466 the sections where they will have the most visual impact.
468 Tuning for the more exact quantizer.
469 Relax rate control for last few frames
471 Limit key frame quantizer for forced key frames.
473 Add simple version of activity masking.
474 make rdmult adaptive for intra in quantizer RDO
475 cap the best quantizer for 2nd order DC
476 change the threshold of DC check for encode breakout
479 Fix crash on Sparc Solaris.
480 Fix counter of fixed keyframe distance
481 ARNR filter pointer update bug fix
482 Fixed use of motion percentage in KF/GF group calc
483 Changed condition for using RD in Intra Mode
484 Fix encoder real-time only configuration.
485 Fix ARM encoder crash with multiple token partitions
486 Fixed bug first cluster timecode of webm file is wrong.
487 Fixed various encoder bugs with odd-sized images
488 vp8e_get_preview fixed when spatial resampling enabled
489 quantizer: fix assertion in fast quantizer path
490 Allocate source buffers to be multiples of 16
491 Fix for manual Golden frame frequency
492 Fix drastic undershoot in long form content
495 2010-10-28 v0.9.5 "Aylesbury"
496 Our first named release, focused on a faster decoder, and a better encoder.
499 This release incorporates backwards-incompatible changes to the
500 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
503 * the -q (quiet) option has been removed, and replaced with
504 -v (verbose). the output is quiet by default. Use -v to see
505 the version number of the binary.
507 * The default behavior is now to write output to a single file
508 instead of individual frames. The -y option has been removed.
509 Y4M output is the default.
511 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
512 options must be specified.
514 $ ivfdec -o OUTPUT INPUT
515 $ vpxdec --i420 -o OUTPUT INPUT
517 * If an output file is not specified, the default is to write
518 Y4M to stdout. This makes piping more natural.
520 $ ivfdec -y -o - INPUT | ...
523 * The output file has additional flexibility for formatting the
524 filename. It supports escape characters for constructing a
525 filename from the width, height, and sequence number. This
526 replaces the -p option. To get the equivalent:
528 $ ivfdec -p frame INPUT
529 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
532 * The output file must be specified with -o, rather than as the
535 $ ivfenc <options> INPUT OUTPUT
536 $ vpxenc <options> -o OUTPUT INPUT
538 * The output defaults to webm. To get IVF output, use the --ivf
541 $ ivfenc <options> INPUT OUTPUT.ivf
542 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
546 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
547 vpxdec supports .webm input
548 vpxdec writes .y4m by default
549 vpxenc writes .webm output by default
550 vpxenc --psnr now shows the average/overall PSNR at the end
551 ARM platforms now support runtime cpu detection
552 vpxdec visualizations added for motion vectors, block modes, references
553 vpxdec now silent by default
554 vpxdec --progress shows frame-by-frame timing information
555 vpxenc supports the distinction between --fps and --timebase
556 NASM is now a supported assembler
557 configure: enable PIC for shared libs by default
558 configure: add --enable-small
559 configure: support for ppc32-linux-gcc
560 configure: support for sparc-solaris-gcc
563 Improve handling of invalid frames
564 Fix valgrind errors in the NEON loop filters.
565 Fix loopfilter delta zero transitions
566 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
567 Build fixes for darwin-icc
570 20-40% (average 28%) improvement in libvpx decoder speed,
572 Rewrite vp8_short_walsh4x4_sse2()
573 Optimizations on the loopfilters.
574 Miscellaneous improvements for Atom
575 Add 4-tap version of 2nd-pass ARMv6 MC filter.
576 Improved multithread utilization
577 Better instruction choices on x86
578 reorder data to use wider instructions
579 Update NEON wide idcts
580 Make block access to frame buffer sequential
581 Improved subset block search
582 Bilinear subpixel optimizations for ssse3.
583 Decrease memory footprint
585 Encoder speed improvements (percentage gain not measured):
586 Skip unnecessary search of identical frames
587 Add SSE2 subtract functions
588 Improve bounds checking in vp8_diamond_search_sadx4()
589 Added vp8_fast_quantize_b_sse2
592 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
593 encoding mode, and up to 60% improvement on very noisy, still
594 or slow moving source video
596 Motion compensated temporal filter for Alt-Ref Noise Reduction
597 Improved use of trellis quantization on 2nd order Y blocks
598 Tune effect of motion on KF/GF boost in two pass
599 Allow coefficient optimization for good quality speed 0.
600 Improved control of active min quantizer for two pass.
601 Enable ARFs for non-lagged compress
605 Disable frame dropping by default
606 Improved multithreaded performance
607 Improved Force Key Frame Behaviour
608 Increased rate control buffer level precision
609 Fix bug in 1st pass motion compensation
610 ivfenc: correct fixed kf interval, --disable-kf
612 Changed above and left context data layout
613 Rework idct calling structure.
614 Removed unnecessary MB_MODE_INFO copies
615 x86: SSSE3 sixtap prediction
616 Reworked IDCT to include reconstruction (add) step
617 Swap alt/gold/new/last frame buffer ptrs instead of copying.
618 Improve SSE2 loopfilter functions
619 Change bitreader to use a larger window.
620 Avoid loopfilter reinitialization when possible
622 Normalize quantizer's zero bin and rounding factors
623 Add trellis quantization.
624 Make the quantizer exact.
625 Updates to ARNR filtering algorithm
626 Fix breakout thresh computation for golden & AltRef frames
627 Redo the forward 4x4 dct
628 Improve the accuracy of forward walsh-hadamard transform
629 Further adjustment of RD behaviour with Q and Zbin.
631 Allow linking of libs built with MinGW to MSVC
632 Fix target auto-detection on mingw32
633 Allow --cpu= to work for x86.
634 configure: pass original arguments through to make dist
635 Fix builds without runtime CPU detection
636 msvs: fix install of codec sources
637 msvs: Change devenv.com command line for better msys support
638 msvs: Add vs9 targets.
639 Add x86_64-linux-icc target
641 Potential crashes on older MinGW builds
642 Fix two-pass framrate for Y4M input.
643 Fixed simple loop filter, other crashes on ARM v6
644 arm: fix missing dependency with --enable-shared
645 configure: support directories containing .o
646 Replace pinsrw (SSE) with MMX instructions
647 apple: include proper mach primatives
648 Fixed rate control bug with long key frame interval.
649 Fix DSO link errors on x86-64 when not using a version script
650 Fixed buffer selection for UV in AltRef filtering
655 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
656 * Speed optimizations
659 * Prevent out-of-bounds accesses on invalid data
660 - Build system updates:
661 * Detect toolchain to be used automatically for native builds
662 * Support building shared libraries
663 * Better autotools emulation (--prefix, --libdir, DESTDIR)
665 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
669 - Initial open source release. Welcome to WebM and VP8!