ffmpeg tips

[aben20807]

Convert flv to mp4 without re-encode

• ref: https://superuser.com/a/860127

ffmpeg -i input.flv -c copy -copyts output.mp4

Convert unplayable mp4 to samsung supported mp4

ffmpeg -n -nostdin -i input.mp4 -c:v h264_nvenc -preset slow -crf 23 -c:a aac -b:a 192k -vf format=yuv420p -movflags +faststart output.mp4

• Document: https://trac.ffmpeg.org/wiki/Encode/H.264
• batch version: https://gist.github.com/aben20807/56d222dae9f4cb7b1dc3f69388e83982

[aben20807]

Convert video to gif

• ref

ffmpeg -y -i input.flv -filter_complex "fps=4,scale=1920:-1:flags=lanczos,split[s0][s1];[s0]palettegen=max_colors=256[p];[s1][p]paletteuse=dither=bayer" output.gif

scale and max_colors impact the output gif's size

[aben20807]

Concat (w/o re-encode - fast)

• ref
• input: A.mp4, B.mp4, C.mp4, ...

ffmpeg -safe 0 -f concat -i <(find . -type f -name '*.mp4' -printf "file '$PWD/%p'\n" | sort) -c copy output.mp4

Concat (w/ re-encode - slow)

• remove -c:v h264_nvenc if there is no GPU driver

ffmpeg -i A.mp4 -i B.mp4 -filter_complex "concat=n=2:v=1:a=1" -c:v h264_nvenc output.mp4

if error when different DAR

[Parsed_concat_4 @ 0x55c55ae4d400] Input link in0:v0 parameters (size 1280x720, SAR 1:1) do not match the corresponding output link in0:v0 parameters (1280x720, SAR 8001:8000)
[Parsed_concat_4 @ 0x55c55ae4d400] Failed to configure output pad on Parsed_concat_4
Error reinitializing filters!

Correct the aspect of one of videos first.

ffmpeg -i tmpA.mp4 -aspect 16:9 -c copy A.mp4

[aben20807]

Convert webp images into jpeg format

• ref

find . -iname '*.webp' -exec bash -c 'ffmpeg -y -hide_banner -loglevel error -i "$1" -q:v 1 -qmin 1 -qmax 1 "${1%.*}.jpg"; printf "$1\n"' _ {} \;

轉換完後刪除原檔

find . -iname '*.webp' -exec bash -c 'ffmpeg -y -hide_banner -loglevel error -i "$1" -q:v 1 -qmin 1 -qmax 1 "${1%.*}.jpg"; printf "$1\n" && rm "$1"' _ {} \;

Using PIL is much faster

• tested in windows powershell with python 3.12.5

# -*- coding: utf8 -*-

# pip install Pillow tqdm
# python -X utf8 .\webp2jpg.py --dir ../

from pathlib import Path
from multiprocessing import Pool, cpu_count
import argparse
import os, sys
from tqdm import tqdm
from PIL import Image

def process(file):
    ori_dir = str(str(file.parent.resolve()).encode("utf8"), encoding="utf8")
    new_dir = ori_dir + "_jpg"
    os.makedirs(new_dir, exist_ok=True)

    ori_file = str(file.resolve())
    new_file = ori_file.replace(file.suffix, ".jpg").replace(ori_dir, new_dir)
    ori_file = str(bytes(ori_file, encoding="utf8"), encoding="utf8")
    new_file = str(bytes(new_file, encoding="utf8"), encoding="utf8")

    if not os.path.isfile(new_file):
        im = Image.open(ori_file).convert("RGB")
        im.save(new_file, "jpeg", subsampling=0, quality=95)

def main():
    parser = argparse.ArgumentParser()
    parser.add_argument("--dir", type=str, required=True)
    parser.add_argument("--count", action="store_true")
    args = parser.parse_args()

    pwd = Path(rf"{args.dir}")
    files = [Path(file) for file in pwd.glob("**/*.webp")]

    if args.count:
        print(f"total: {len(files)}")
        sys.exit(0)

    with Pool(cpu_count()) as p:
        with tqdm(total=len(files), ascii=True, ncols=60) as pbar:
            for _ in p.imap_unordered(process, files):
                pbar.update()

    print(f"total: {len(files)}")

if __name__ == "__main__":
    main()

[aben20807]

Reduce file size

• -crf N: N: 0–51, where 0 is lossless, 23 is the default, and 51 is worst quality possible
• -b 800k: limit the bitrate (optional)

ffmpeg -i input.flv -c:v h264_nvenc -crf 30 -c:a aac -b:a 192k -vf format=yuv420p -movflags +faststart output.mp4

[aben20807]

Combine video and audio

ffmpeg -i video.mp4 -i audio.m4a -c copy -map 0:v:0 -map 1:a:0 output.mp4

optional

 yt-dlp -F {youtube url}