libaom v3.14.0 Pinkie Pie Released

[Wan-Teh Chang]

libaom v3.14.0 Pinkie Pie has been released. The source code of the
release can be checked out from the git repository[1] using the release
tag v3.14.0. Alternatively, the release tarball[2], along with the
signature file[3], can be downloaded.

Special thanks to the Arm team for their contributions to the
optimizations in this release.

Thanks.

[1] https://aomedia.googlesource.com/aom
[2] https://storage.googleapis.com/aom-releases/libaom-3.14.0.tar.gz
[3] https://storage.googleapis.com/aom-releases/libaom-3.14.0.tar.gz.asc

Release Notes:

2026-05-12 v3.14.0
This release is ABI compatible with the last release.

This release includes significant compression efficiency and
perceptual quality improvements for layered image encoding, in
addition to bug fixes.

The `build` directory in the source tree was removed and the original
build/cmake directory was moved up one level. If you use the CMake
toolchain files in the original build/cmake/toolchains directory,
they are now in the cmake/toolchains directory.

The unsupported aom_img_fmt_t enum constants AOM_IMG_FMT_AOMYV12 and
AOM_IMG_FMT_AOMI420 are deprecated and will be removed in a future
release. Do not use them.

The new codec controls AV1E_SET_EXTERNAL_RATE_CONTROL (including the
new aom_rc_funcs_t struct and the new "aom/aom_ext_ratectrl.h" and
"aom/aom_tpl.h" headers) and AV1E_GET_GOP_INFO (including the new
aom_gop_info_t struct) have the experimental API status and are NOT
part of the stable API. Applications that need the backward
compatibility of the stable API must not use
AV1E_SET_EXTERNAL_RATE_CONTROL and AV1E_GET_GOP_INFO.

- New Features
* Tuning modes AOM_TUNE_IQ and AOM_TUNE_SSIMULACRA2 now work with
inter-frame encoding modes (good-quality and realtime), enabling
efficient layered image encoding.
* Variance Boost (deltaq-mode 6) support has been extended to
include good-quality and realtime modes.
* New value 2 for use_fixed_qp_offsets: disallow the application of
per-frame QP offsets, enabling full control of each frame's QP, by
adjusting rc_cfg.cq_level between each encoded frame.
* Low complexity decode mode (the encoder optimizations that produce
encoded bitstreams that can be decoded at low computational
complexity) now supports good-quality encoding (speed 1 to 3) for
not only vertical videos (608p to 1080p) but also horizontal
videos (720p to 1080p).
* Realtime encoding with non-zero lookahead is supported, for
!REALTME_ONLY_CONFIG.
* New AOM_EFLAG_FREEZE_INTERNAL_STATE flag for aom_codec_encode() to
encode a frame without updating encoder state such as reference
buffers, CDF tables, and rate control state. Useful for
speculative encoding.
* New aom_matrix_coefficients_t enum constants:
- AOM_CICP_MC_IPT_C2 (15): IPT-C2
- AOM_CICP_MC_YCGCO_RE (16): YCgCo-Re
- AOM_CICP_MC_YCGCO_RO (17): YCgCo-Ro
* New aom_metadata_insert_flags_t enum constants for adding
layer-specific metadata OBUs:
- AOM_MIF_NON_KEY_FRAME_LAYER_SPECIFIC: Adds layer-specific
metadata if it's not a keyframe
- AOM_MIF_KEY_FRAME_LAYER_SPECIFIC: Adds layer-specific metadata
only if it's a keyframe
- AOM_MIF_ANY_FRAME_LAYER_SPECIFIC: Adds layer-specific metadata
to any type of frame
* Experimental: New codec control AV1E_SET_EXTERNAL_RATE_CONTROL to
enable an external rate control library. Not part of the stable
API.
* Experimental: New codec control AV1E_GET_GOP_INFO to get the GOP
structure from the encoder. Not part of the stable API.
* New codec control AOME_SET_VALIDATE_HBD_INPUT to validate high
bitdepth (HBD) input and ensure that every pixel is within the
valid range. The corresponding option for aom_codec_set_option()
is "validate-hbd-input".
* New codec control AOMD_SET_FRAME_SIZE_LIMIT to set the maximum
frame size for a decoder

- Compression Efficiency Improvements
* When encoding 2-layered images with AOM_TUNE_IQ vs. AOM_TUNE_SSIM:
- Good-quality mode: up to 15% gains at similar SSIMULACRA 2
scores.
- Realtime mode: up to 30% gains at similar SSIMULACRA 2 scores.
* Enable screen detection mode 2 (anti-aliased text and graphics
aware) in all-intra mode.
* Some minor quantization matrix formula and delta chroma q tweaks
for the AOM_TUNE_SSIMULACRA2 tuning mode, with up to 0.2%
SSIMULACRA 2 score efficiency gains.
* Re-tune and re-work encoder features, achieving coding gains at
each speed level and better coding efficiency vs complexity
tradeoff (an estimated 20-30% encoder time reduction with about
1-5% vmaf gains).
* RTC screen: improvements to quality and scroll detection for high
resolutions.

- Perceptual Quality Improvements
* AOM_TUNE_IQ improves the visual quality of layered image encoding
compared to AOM_TUNE_SSIM.
* Adaptive sharpness: tweak loop filter sharpness threshold to
improve visual quality at QPs 29 and 30.
* Improve encoder algorithms in loop restoration, CDEF, temporal
filter and rate control, delivering better visual quality.

- Speedups
* Make Adaptive CDEF more decoder friendly by limiting CDEF
application to only regions where high filter strengths are
needed.
* Encoder complexity is reduced significantly, especially at speed
1 to 3.
* Further AArch64 SIMD optimization of filtering algorithms,
particularly using Armv8.6 Neon I8MM.
* Additional tweaks to AArch64 Neon paths for quantization, SAD,
subpel variance and intra-predictors.

- Bug Fixes
* Commit 4cc0867f: fix multithreading crashes when loop restoration
filtering is used by the encoder.
* Commit 74aaa1ef: fix AOM_SCALING_MODE crashes when compound mask
prediction modes are used by the encoder.
* b:502030569: Avoid assertion failure in `read_uncompressed_header`
triggered by fuzzed input
* b:503197490: Fix an integer overflow in target bits calculation
* b:503691210: Avoid reference to uninitialized member variables for
multi workers.
* b:504317456: Handle buffer pointer in LAP mode to avoid overflow
* b:502133197: Avoid a bitread assertion failure in
get_av1config_from_obu()