初探 FFMPEG 视频转高质量 GIF

1. 缘起

2. 透明度优化

$ ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1" ffmpeg_default.gif
$ ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1" -gifflags -transdiff ffmpeg_default_no_transdiff.gif
$ ls -lh ffmpeg_default*gif
-rw-r--r--  1 chenyang  staff   1.4M 10  6 01:46 ffmpeg_default.gif
-rw-r--r--  1 chenyang  staff   1.5M 10  6 01:45 ffmpeg_default_no_transdiff.gif

$ mkdir ffmpeg_default_frames && convert ffmpeg_default.gif ffmpeg_default_frames/%d.png
$ mkdir ffmpeg_default_frames_no_transdiff && convert ffmpeg_default_no_transdiff.gif ffmpeg_default_frames_no_transdiff/%d.png

3. 抖动(Dithering)

3.1 调色板

# 使用默认参数, 等效于 ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1:flags=lanczos,palettegen=max_colors=256:reserve_transparent=on" palette.png
$ ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1:flags=lanczos,palettegen" palette.png

# 只用 100 种颜色 (最后一种为透明)
$ ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1:flags=lanczos,palettegen=max_colors=100" palette_100color.png

3.2 抖动算法

for each y from top to bottom do
    for each x from left to right do
        oldpixel := pixels[x][y]
        newpixel := find_closest_palette_color(oldpixel)
        pixels[x][y] := newpixel
        quant_error := oldpixel - newpixel
        pixels[x + 1][y    ] := pixels[x + 1][y    ] + quant_error × 7 / 16
        pixels[x - 1][y + 1] := pixels[x - 1][y + 1] + quant_error × 3 / 16
        pixels[x    ][y + 1] := pixels[x    ][y + 1] + quant_error × 5 / 16
        pixels[x + 1][y + 1] := pixels[x + 1][y + 1] + quant_error × 1 / 16

ffmpeg -i demo.mp4 -i palette.png -lavfi "fps=10,scale=480:-1:flags=lanczos [x]; [x][1:v] paletteuse=dither=bayer:bayer_scale=2" out.gif

dither = none 时出现了明显的颜色带效应, 因为相邻区域颜色相近, 都划分到了同一个颜色.

4. 高质量 GIF 生成

ffmpeg -i demo.mp4 -vf "fps=10,scale=480:-1:flags=lanczos,split[split1][split2];[split1]palettegen[pal];[split2][pal]paletteuse=dither=none" demo_dither_none.gif

set -e

# global filter
fps=10
scale=480:-1
interpolation=lanczos

# for palettegen
max_colors=256  # up to 256
reserve_transparent=on
stats_mode=full  # chosen from [full, diff, single]

# for paletteuse
dither=bayer  # chosen from [bayer, heckbert, floyd_steinberg, sierra2, sierra2_4a, none]
bayer_scale=3  # [0, 5]. only works when dither=bayer. higher means more color banding but less crosshatch pattern and smaller file size
diff_mode=rectangle  # chosen from [rectangle, none]
new=off  # when stats_mode=single and new=on, each frame uses different palette

ffmpeg -i $1 -vf "fps=$fps,scale=$scale:flags=$interpolation,split[split1][split2];[split1]palettegen=max_colors=$max_colors:reserve_transparent=$reserve_transparent:stats_mode=$stats_mode[pal];[split2][pal]paletteuse=dither=$dither:bayer_scale=$bayer_scale:diff_mode=$diff_mode:new=$new" -y $2

$ gifsicle --sinfo ezgif_default.gif
* ezgif_default.gif 32 images
  logical screen 480x270
  global color table [256]
  background 0
  loop forever
  + image #0 480x270 transparent 0
    compressed size 68020
    comment Created with ezgif.com video to GIF converter
    local color table [256]
    disposal asis delay 0.10s
  + image #1 480x270 transparent 0
    compressed size 74188
    disposal asis delay 0.10s
	... 省略 ...

$ gifsicle --sinfo new_res.gif
* new_res.gif 32 images
  logical screen 480x270
  global color table [256]
  background 0
  loop forever
  + image #0 480x270 transparent 0
    compressed size 53187
    local color table [256]
    disposal asis delay 0.10s
  + image #1 480x270 transparent 0
    compressed size 58854
    local color table [256]
    disposal asis delay 0.10s
	... 省略 ...

High quality GIF with FFmpeg

About two years ago, I tried to improve the support of the GIF encoding in FFmpeg to make it at least decent. This notably led to the addition of the transparency mechanism in the GIF encoder. While this is not always optimal depending on your source, it is in the most common cases.

http://blog.pkh.me/p/21-high-quality-gif-with-ffmpeg.html

cassidy.codes

A few months back I was working on an iOS app that delivered GIFs to users. The idea was that someone would visit one of our 3D photo booths, and get an animated avatar created. We could then render animation frames to PNGs using our custom software.

https://cassidy.codes/blog/2017/04/25/ffmpeg-frames-to-gif-optimization/

Dither - Wikipedia

Dither is an intentionally applied form of noise used to randomize quantization error, preventing large-scale patterns such as color banding in images. Dither is routinely used in processing of both digital audio and video data, and is often one of the last stages of mastering audio to a CD.

https://en.wikipedia.org/wiki/Dither

压缩gif的正确姿势

有时候，我们有压缩gif的需求--例如为了优化网页加载速度。在unix环境下，可以使用ImageMagick工具。 压缩gif使用-layers方法 Optimize ： convert source.gif -layers Optimize dest.gif ...

https://segmentfault.com/a/1190000000436384

How to make GIFs with FFMPEG

To follow along, download media files here: https://github.com/cyburgee/ffmpeg-guide If you've worked with media encoding in the past decade it's likely that you've come across FFmpeg. For those of you who are unfamiliar, in their words: "FFmpeg is the leading multimedia framework, able to decode, encode, transcode, mux, demux, stream, filter and play pretty much anything that humans and machines have created.

https://engineering.giphy.com/how-to-make-gifs-with-ffmpeg/

[FFMpeg开发]视频转高质量GIF优化方案(接近ps生成效果)，从原理剖析_没有技术的菜逼的博客-CSDN博客_ffmpeg 高质量gif

虽然此前有人发过了，但是这个博主没有分析原理并且没有提炼出来。不适合开发者学习。 所以我只是进行二次优化，原文 高质量视频转gif ...

https://blog.csdn.net/qq_40212938/article/details/113060064