Intra vs Inter Frame
Also known as: i frame vs p frame, intra inter coding, keyframe vs delta frame, intra vs inter frame
Intra (I) frames are coded as standalone images, like a JPEG. Inter (P/B) frames store only the changes from nearby frames, so they are far smaller. This temporal redundancy is why video files shrink dramatically compared to storing each frame as a photo.
- I-frames are self-contained (like a JPEG); P/B-frames store only changes from reference frames.
- A run of frames from one I-frame to the next is called a GOP (Group of Pictures).
- Temporal redundancy between inter frames is why 1 second of video is far smaller than 30 photos.
How intra and inter frames differ
An intra-coded frame (I-frame, also called a keyframe) is compressed using only information inside that one frame, similar to a standalone JPEG. It can be decoded on its own with no reference to any other frame, which makes it a seek point and an error-recovery anchor.
An inter-coded frame stores motion vectors and residuals that describe how the picture differs from one or more reference frames. P-frames (predicted) reference an earlier frame; B-frames (bi-directional) reference both earlier and later frames. Because most pixels barely change between consecutive frames, these deltas are tiny compared to a full image.
Why this makes video so much smaller than photos
A 30 fps clip has 30 images per second, yet a single second of video is usually far smaller than 30 separate photos. The codec exploits temporal redundancy: it sends one I-frame, then a long run of P/B-frames that only encode movement and small changes.
The sequence between two I-frames is a GOP (Group of Pictures). Longer GOPs and more B-frames mean smaller files but slower seeking and heavier decoding. Codecs like H.264, HEVC, and AV1 all rely on this intra/inter split; the main difference between them is how cleverly they predict and quantize the inter frames.