Video Compression Guide: Reduce Size, Not Quality

Compression is not magic—it is applied math that throws away information humans rarely notice. Understanding codecs, bitrates, and rate control modes lets you shrink files deliberately instead of clicking “small” and hoping. This guide focuses on theory and practical encoding decisions, not inbox limits (see our email-specific article for attachment workflows).

Lossless vs lossy compression

Lossless schemes (like FFV1 or HuffYUV in archival contexts) shrink data without altering pixels after decode. They are used in mastering and medical imaging, not for YouTube delivery. Lossy codecs—H.264/AVC, HEVC/H.265, VP9, AV1—exploit spatial and temporal redundancy: similar blocks across frames are stored once, and motion vectors describe how blocks move. The encoder chooses what to discard based on a quality target or bitrate cap. Every lossy pass is destructive; re-compressing an already compressed file compounds artifacts (“generation loss”).

Containers vs codecs

MP4, MKV, and WebM are containers; H.264 and AAC inside MP4 are codecs. Changing containers without re-encoding (remuxing) does not shrink video. Real savings come from re-encoding with a more efficient codec or lower bitrate target. The free video compressor exposes presets that map to sensible CRF and resolution pairs so you do not need to memorize ffmpeg flags.

Codec families in 2024–2026

H.264 remains the compatibility king: hardware decode everywhere, fast encodes, good quality around 5–12 Mbps for 1080p. HEVC (H.265) cuts bitrate roughly 30–50% at the same perceptual quality but encodes slower and may lack hardware support on older PCs. VP9 and AV1 dominate web platforms; AV1 is increasingly default on YouTube and Netflix, with excellent efficiency at the cost of encode time. For browser-first workflows, H.264 in MP4 is still the safest handoff; use AV1 when the platform ingests it natively and you have time to encode.

Bitrate ladders and resolution

A bitrate ladder is a set of rungs (resolution + bitrate + codec) for adaptive streaming. Even for single-file delivery, thinking in ladders prevents over- or under-shooting:

Resolution	Frame rate	H.264 target (VBR)	HEVC target (VBR)
3840×2160 (4K)	30	35–45 Mbps	20–28 Mbps
1920×1080	30	8–12 Mbps	5–8 Mbps
1280×720	30	4–6 Mbps	2.5–4 Mbps
854×480	30	1.5–2.5 Mbps	0.8–1.5 Mbps

Talking-head footage can sit at the low end; sports and confetti need the high end. Downscaling 4K to 1080p often beats brutal bitrate slashing on 4K itself.

CRF, VBR, and two-pass encoding

CRF (Constant Rate Factor) targets quality, not file size. For H.264, CRF 18–20 is visually transparent for most content; 23 is a common web default; 28+ shows obvious softness. 1-pass VBR hits an average bitrate with peaks for motion. 2-pass VBR analyzes complexity first, then allocates bits—better for DVD-era constraints, less critical for short web clips. When you need a hard ceiling (CDN budget), combine max bitrate caps with buffer size; when you need predictable size, use CRF and trim duration first with trim video.

Audio is half the experience

AAC-LC at 128–192 kbps stereo is standard; speech podcasts can use 96 kbps mono. Do not crush video bitrate to save audio—bad audio makes “high quality” video feel cheap. Extract audio-only with video to MP3 when video is unnecessary.

GOP structure and seeking behavior

Group of pictures (GOP) length affects how quickly players seek and how big each keyframe is. Long GOPs (e.g., 4 seconds) improve compression but make scrubbing less responsive in editors. Short GOPs help streaming start faster. When you compress for web, a GOP around half the frame rate (about 15 frames at 30 fps) is a reasonable default. B-frames improve efficiency but can confuse some old hardware decoders—most web targets tolerate them in H.264.

Chroma subsampling and sharp text

4:2:0 subsampling, standard in H.264 and HEVC delivery, reduces color resolution compared to luma. Screen recordings with red UI on grey backgrounds may show fringing after heavy compression. Slightly higher bitrate or 4:2:2 in intermediate mezzanine files (ProRes) before a 4:2:0 web encode preserves UI clarity. For slide decks, prefer CRF 20–22 instead of starving bitrate to hit an arbitrary megabyte number.

Practical encoding checklist

1. Start from the highest-quality master you have; never up-scale.
2. Trim dead sections before encoding to save bits on content, not silence.
3. Pick codec by audience device (H.264 default; HEVC/AV1 when supported).
4. Choose CRF for archival-friendly web posts; choose capped VBR for strict size targets.
5. Preview on a phone, not only a 27" monitor—blocking artifacts show on small screens.

Need a different container after compression? Use the video converter. For platform-specific dimensions, pair compression with crop before encode to avoid padding black bars into the bitrate budget. See all features for speed, mute, and loop utilities that should run on the master before the final encode.

Scene complexity and perceptual quality

Not every frame deserves the same number of bits. Static interviews compress easily; confetti, rain, and handheld night footage need higher peaks. Two-pass VBR historically allocated bits to hard scenes first; modern single-pass CRF with good lookahead closes much of that gap. If your footage is mostly talking head with a static background, bias toward the lower end of the bitrate ladder; for sports or concerts, bias high or accept larger files. Always watch the output on the device your audience uses—phones hide mild blocking that large TVs reveal.

Frequently asked questions

What CRF should I use for YouTube uploads?

YouTube re-encodes everything. Upload a high-quality H.264 MP4 (CRF 18–20 at 1080p) so the platform has headroom; avoid uploading already blocky CRF 28 files.

Is HEVC worth it for email or chat apps?

Often no—compatibility beats efficiency. HEVC shines for personal archives and Apple-centric workflows, not universal sharing.

Why does my compressed file look worse in shadows?

Low bitrates crush chroma and fine grain. Raise bitrate slightly, use a slower preset, or denoise source footage before encode.

Can I compress without re-encoding?

Only remuxing or trimming via stream copy avoids re-encode. True size reduction requires a new lossy encode or a more efficient codec.

Conclusion: Compression is the art of spending bits where eyes notice. Learn your codec, pick a sensible ladder, use CRF for quality-first exports, and trim before you encode. The compressor automates the hard choices while you keep creative control over the master.