What Is AVIF, and Why Is It So Much Smaller?
AVIF files routinely come in at half the size of an equivalent JPG, and noticeably smaller than WebP too, at the same visual quality. The reason isn't "newer format, better numbers." AVIF is built on a genuinely different compression foundation than JPG and WebP, borrowed wholesale from video, and that foundation is what actually produces the savings.
It's a video codec still image
AVIF stands for AV1 Image File Format. It isn't a new image compression scheme, it's a single still frame encoded using AV1, the open video codec developed for streaming video. AV1 has an intra-frame compression mode, the part of a video codec responsible for compressing one frame with no reference to neighboring frames, and that's exactly the mode AVIF reuses for still images. JPG, by contrast, is built on discrete cosine transform (DCT) compression designed in the late 1980s. WebP's lossy mode is based on the intra-frame compression from VP8, itself roughly a decade older than AV1. AVIF is a full codec generation newer than both.
Where the real savings come from
Two specific improvements in AV1 do most of the work. First, larger and more flexible prediction block sizes: JPG divides an image into fixed 8x8 pixel blocks and compresses each independently, which is exactly why heavily compressed JPGs show visible blocky squares. AV1 can use much larger blocks (up to 128x128) and choose block shapes adaptively based on the image content, so a large smooth area, a sky, a wall, skin, gets described far more efficiently than JPG's fixed grid ever could.
Second, better handling of smooth gradients specifically. JPG's DCT-based approach struggles to represent a slow, continuous color transition without introducing visible steps, which is the technical cause of banding, the stripes you see in a low-quality JPG sky or gradient background. AV1's prediction modes represent gradients more directly, which is why AVIF holds up noticeably better than JPG at aggressive compression settings specifically on photos with large smooth regions, not just "generally better," but better in a way traceable to a specific mechanism.
The real trade-off: encoding is slow
None of this is free. AV1's extra flexibility, larger block search space, more prediction modes to evaluate, means an encoder has far more decisions to make per image than a JPG or WebP encoder does. AVIF encoding is meaningfully heavier computationally, often several times slower than encoding the same image to JPG or WebP. This matters if you're batch-converting a large folder of images, the encode time adds up in a way it wouldn't with older formats. Decoding an AVIF, actually viewing it, is comparatively fast, the asymmetry is specific to encoding, not viewing. Kit-Bin's converters run the encode in your browser, so that computational cost happens locally on your own device rather than requiring any server, but it's still real CPU time worth expecting on larger batches.
Using it
JPG to AVIF and AVIF to JPG both run entirely in your browser. If you just want a smaller file without changing format, Compress Image is the more direct tool.
Related guides
If an AVIF file refuses to open somewhere, that's usually a decoder support gap rather than a broken file. See Why Won't My AVIF File Open?. AVIF also supports a full alpha channel, same as PNG or WebP — see which image formats support transparency for how it compares.