Pixel Formats — yuv420p and Why It Matters

The chroma-subsampling alphabet soup

A pixel format string like yuv420p10le packs four pieces of information:

yuv — color space (vs rgb).
420 — chroma subsampling. 4:4:4 = full color, 4:2:2 = half horizontal chroma, 4:2:0 = quarter chroma (¼ the chroma data of 4:4:4).
p — planar layout (separate Y, U, V planes).
10le — 10-bit little-endian (vs the default 8-bit).

What you'll actually see

yuv420p — 8-bit 4:2:0. The H.264 web/mobile default. Use it.
yuv420p10le — 10-bit 4:2:0. HEVC HDR, finer gradients, slight quality bump.
yuv422p10le — 10-bit 4:2:2. ProRes 422 default.
yuva444p10le — 10-bit 4:4:4 with alpha. ProRes 4444.

The trap

If you decode a 4:2:0 source and re-encode without specifying -pix_fmt, FFmpeg picks the encoder's preferred format. libx264 will sometimes emit 4:4:4 by accident if the source was unusual, and that file won't play on iPhone, Safari, or most TVs. Always explicitly set -pix_fmt yuv420p for compatibility outputs.

Code

Force the right pixel format·bash

# 8-bit 4:2:0 — universal compatibility
ffmpeg -i in.mp4 -c:v libx264 -crf 20 -pix_fmt yuv420p out.mp4

# 10-bit 4:2:0 — slightly better gradients (HEVC, browsers OK)
ffmpeg -i in.mp4 -c:v libx265 -crf 24 -pix_fmt yuv420p10le -tag:v hvc1 out.mp4

# 10-bit 4:2:2 — for ProRes editing master
ffmpeg -i in.mp4 -c:v prores_ks -profile:v 3 -pix_fmt yuv422p10le master.mov

Inspect what you ended up with·bash

# What pixel format did the output land in?
ffprobe -v error -select_streams v:0 \
  -show_entries stream=pix_fmt,bits_per_raw_sample \
  -of default=noprint_wrappers=1 out.mp4
# pix_fmt=yuv420p
# bits_per_raw_sample=8

Pixel Formats — yuv420p and Why It Matters

The chroma-subsampling alphabet soup

What you'll actually see

The trap

Code

External links

Exercise

Progress

Comments 0