pyavif 0.0.1

AVIF bindings for Python with NumPy support

pyavif

AVIF bindings for Python with NumPy support.

Installation

pip install numpy
pip install .

Basic Usage

Decoding

from pyavif import Decoder, DecoderCodec

# Create decoder with default codec (DAV1D - fastest decoder)
decoder = Decoder()
decoder.init("image.avif", decoder_threads=1)
image = decoder.get_image(0)  # Returns numpy array (H, W, C)

# Or specify a different codec
decoder = Decoder()
decoder.init("image.avif", decoder_threads=1, codec=DecoderCodec.AOM)

Encoding

import numpy as np
from pyavif import Encoder, EncoderOptions, EncoderCodec

# Create encoder
opts = EncoderOptions()
# Use alternative codec
opts.codec = EncoderCodec.RAV1E
encoder = Encoder("output.avif", width=256, height=256, channels=3, depth=8, options=opts)

# Add frame and finish
image = np.zeros((256, 256, 3), dtype=np.uint8)
encoder.add_frame(image)

# finish when done
encoder.finish()

Advanced

Batch Encoding

from pyavif import BatchEncoder, EncoderOptions

opts = EncoderOptions()
output_paths = ["out1.avif", "out2.avif", "out3.avif"]
batch_encoder = BatchEncoder(output_paths, opts)

# Add frames for all files at once
frames = [image1, image2, image3]  # Same-length list as output_paths
batch_encoder.add_image_batch(frames, duration=1)

batch_encoder.finish_all()

Encoder Options

from pyavif import EncoderOptions, EncoderCodec

opts = EncoderOptions()
opts.codec = EncoderCodec.RAV1E  # RAV1E, AOM (default: AOM)
opts.quality = 85               # 0-100 (default: 80)
opts.speed = 6                  # 0-10, higher = faster (default: AVIF_SPEED_DEFAULT)
opts.quality_alpha = 100        # Alpha channel quality (default: AVIF_QUALITY_LOSSLESS)
opts.timescale = 30             # Animation timescale in fps (default: 30)
opts.max_threads = 4            # Thread count (default: 1)

Building libavif

We ship prebuilt static libavif artifacts in lib/ to keep installs simple. If you need to rebuild them, use the scripts below. They clone libavif at v1.3.0, build static libraries with RAV1E, DAV1D, AOM, and libyuv, and refresh headers in lib/include/avif.

• Linux (manylinux_2_28, glibc 2.28+): bash scripts/build_libavif_linux.sh (Docker required)
• macOS: bash scripts/build_libavif_macos.sh (requires CMake, Rust, nasm, meson, ninja)
• Windows: pwsh scripts/build_libavif_windows.ps1 (requires CMake, Rust, NASM, Perl, Meson/Ninja)

Set LIBAVIF_VERSION to override the libavif tag.

CI Wheels

GitHub Actions builds wheels via cibuildwheel using .github/workflows/build_wheels.yml. Trigger the workflow manually or on a version tag (for example v0.0.1).