vid_convert/vid_convert.py

#!/usr/bin/env python3
# -*- coding: utf-8 -*-

import json
import subprocess
import logging


def get_infos(file):
    '''
    Cette fonction extrait les informations du film à l'aide de ffprobe et les stocke
    dans un dictionnaire pour une utilisation ultérieure.

    -> http://ffmpeg.org/ffprobe.html
    '''
    v_infos = {
            'index': None,
            'height': None,
            'width': None,
            'color_primaries': None,
            'color_space': None,
            'color_transfer': None,
            'pix_fmt': None,
            'display_aspect_ratio': None,
            }
    a_infos = {}
    v_infos_cmd = f'ffprobe -v quiet -print_format json -show_format -show_streams -show_frames -read_intervals "%+#1" -select_streams v {file}'
    v_infos_raw = subprocess.getoutput(v_infos_cmd)
    a_infos_cmd = f"ffprobe -v quiet -print_format json -show_format -show_streams -select_streams a {file}"
    a_infos_raw = subprocess.getoutput(a_infos_cmd)
    full_v_infos = json.loads(v_infos_raw)
    full_a_infos = json.loads(a_infos_raw)
    v_stream = full_v_infos['streams'][0]
    for prop in v_infos.keys():
        try:
            v_infos.update({prop: v_stream[prop]})
        except KeyError:
            pass
    try:
        v_infos.update({'side_data_list': full_v_infos['frames'][0]['side_data_list']})
    except KeyError:
        pass
    a_infos = []
    for a_stream in full_a_infos['streams']:
        a_stream_infos = {
                'index': a_stream['index'],
                'channels': a_stream['channels'],
                'channel_layout': a_stream['channel_layout'],
                'language': a_stream['tags']['language'],
                'title': a_stream['tags']['title']}
        a_infos.append(a_stream_infos)
    duration = subprocess.getoutput(f"ffprobe -v quiet -print_format json -show_format {file}")
    duration = json.loads(duration)
    duration = float(duration['format']['duration'])
    hdr10_v_cmd = f'ffmpeg -loglevel panic -i {file} -c:v copy -vbsf hevc_mp4toannexb -f hevc - | hdr10plus_parser -o metadata.json --verify -'
    hdr10_v_raw = subprocess.getoutput(hdr10_v_cmd)
    # logging.debug(hdr10_v_raw)
    if 'metadata detected' in hdr10_v_raw:
        hdr10_cmd = f'ffmpeg -loglevel panic -i {file} -c:v copy -vbsf hevc_mp4toannexb -f hevc - | hdr10plus_parser -o /tmp/{file}_hdr10_metadata.json -'
        v_infos.update({'hdr10': True, 'hdr10_metdata': f'/tmp/{file}_hdr10_metadata.json'})
    infos = {'duration': duration, 'video': v_infos, 'audio': a_infos}
    logging.debug("Informations du film : \n" + json.dumps(infos, indent=True))
    return infos


def is_interlaced(file, infos):
    '''
    Cette fonction detecte si la vidéo est entrelacée.
    -> https://fr.wikipedia.org/wiki/Entrelacement_(vid%C3%A9o)
    '''
    duration_tier = int(infos['duration'] / 3)
    command = f"ffmpeg -loglevel info -ss {duration_tier} -t {duration_tier} -i {file} -an -filter:v idet -f null -y /dev/null"
    result = subprocess.getoutput(command)
    for line in result.splitlines():
        if "Multi" in line:
            TFF = int(line.split('TFF:')[1].split()[0])
            BFF = int(line.split('BFF:')[1].split()[0])
            Progressive = int(line.split('Progressive:')[1].split()[0])
    try:
        pct = ((TFF + BFF) / (TFF + BFF + Progressive)) * 100
        pct = round(pct)
    except ZeroDivisionError:
        pct = 100
    if pct > 10:
        logging.debug("Vidéo entrelacée à {pct}%")
        return True
    else:
        logging.debug("Vidéo non entrelacée")
        return False


def cropping(file, infos):
    '''
    Cette fonction detecte les bandes inutiles de la vidéo
    '''
    logging.debug("Détection de la taille de l'image...")
    duration_tier = int(infos['duration'] / 3)
    command = f"ffmpeg -loglevel info -i {file} -ss {duration_tier} -t {duration_tier} -an -f null -vf cropdetect -y /dev/null"
    cropsize = subprocess.getoutput(command).splitlines()[-3].split()[-1]
    logging.debug(f"Paramètre de découpe : {cropsize}")
    return cropsize


def volume_audio(file, infos):
    '''
    Cette fonction ajuste le volume vers 0dB
    '''
    volumes = {}
    for piste_audio in infos['audio']:
        piste = piste_audio['index']
        command = f"ffmpeg -loglevel info -i {file} -map 0:{piste} -af volumedetect -f null -y /dev/null"
        volumedetect = subprocess.getoutput(command)
        for line in volumedetect.splitlines():
            if "max_volume" in line:
                volume = line.split()[-2]
                volume = f"{str(-float(volume))}dB"
        logging.debug(f"Ajustement du volume de la piste {piste} : {volume}")
        volumes.update({piste: volume})
    return volumes


def stabilization(file):
    '''
    Cette fonction permet de stabiliser l'image,
    par exemple quand filmé au smartphone.
    '''
    cmd_stab = f'ffmpeg -i {file} -vf vidstabdetect=shakiness=10:accuracy=10:result="/tmp/vidstab.trf" -f null - '
    subprocess.getoutput(cmd_stab)


def convert_audio(file, track, volume_adj, channels, channel_layout, language, title):
    bitrate = f'{64*channels}k'
    if channel_layout == "5.1(side)":
        channel_layout = "5.1"
    codec = 'libopus'
    metadatas = f'-metadata language="{language}" -metadata title="{title}"'
    command = f'ffmpeg -loglevel error -i {file} -map 0:{track} -vn -sn -c:a {codec} -b:a {bitrate} -mapping_family 1 -filter:a volume={volume_adj},aformat=channel_layouts={channel_layout} -y {file}_audio_{track}.mka'
    logging.debug(command)
    result = subprocess.getoutput(command)
    logging.info(result)


def convert_video(file, infos, start, crop, crf):
    str_start = "{:05d}".format(start)
    output = f'{file}_video_t{str_start}.mkv'
    fmt = infos['video']['pix_fmt']
    track = infos['video']['index']
    codec = 'libx265 -preset slow'
    hdr = ''
    if 'side_data_list' in infos['video'].keys():
        try:
            light_level = f"{infos['video']['side_data_list'][1]['max_content']},{infos['video']['side_data_list'][1]['max_average']}"
            color_primaries = infos['video']['color_primaries']
            color_transfer = infos['video']['color_transfer']
            color_space = infos['video']['color_space']
            green_x = infos['video']['side_data_list'][0]['green_x'].split('/')
            green_x = int(int(green_x[0])*(int(green_x[1])/50000))
            green_y = infos['video']['side_data_list'][0]['green_y'].split('/')
            green_y = int(int(green_y[0])*(int(green_y[1])/50000))
            green = f'G\({green_x},{green_y}\)'
            blue_x = infos['video']['side_data_list'][0]['blue_x'].split('/')
            blue_x = int(int(blue_x[0])*(int(blue_x[1])/50000))
            blue_y = infos['video']['side_data_list'][0]['blue_y'].split('/')
            blue_y = int(int(blue_y[0])*(int(blue_y[1])/50000))
            blue = f'B\({blue_x},{blue_y}\)'
            red_x = infos['video']['side_data_list'][0]['red_x'].split('/')
            red_x = int(int(red_x[0])*(int(red_x[1])/50000))
            red_y = infos['video']['side_data_list'][0]['red_y'].split('/')
            red_y = int(int(red_y[0])*(int(red_y[1])/50000))
            red = f'R\({red_x},{red_y}\)'
            white_point_x = infos['video']['side_data_list'][0]['white_point_x'].split('/')
            white_point_x = int(int(white_point_x[0])*(int(white_point_x[1])/50000))
            white_point_y = infos['video']['side_data_list'][0]['white_point_y'].split('/')
            white_point_y = int(int(white_point_y[0])*(int(white_point_y[1])/50000))
            white_point = f'WP\({white_point_x},{white_point_y}\)'
            min_luminance = infos['video']['side_data_list'][0]['min_luminance'].split('/')
            min_luminance = int(int(min_luminance[0])*(int(min_luminance[1])/10000))
            max_luminance = infos['video']['side_data_list'][0]['max_luminance'].split('/')
            max_luminance = int(int(max_luminance[0])*(int(max_luminance[1])/10000))
            luminance = f'L\({max_luminance},{min_luminance}\)'
            master_display = green + blue + red + white_point + luminance
            hdr = f'-x265-params hdr-opt=1:repeat-headers=1:colorprim={color_primaries}:transfer={color_transfer}:colormatrix={color_space}:master-display={master_display}:max-cll={light_level}'
        except:
            logging.debug("Aucune information HDR")
    command = f'ffmpeg -loglevel error -i {file} -map 0:{track} -ss {start} -t 300 -an -sn -c:v {codec} {hdr} -crf {crf} -pix_fmt {fmt} -filter:v {crop} -y {output}'
    logging.debug(command)
    result = subprocess.getoutput(command)
    logging.info(result)


if __name__ == '__main__':
    import argparse
    parser = argparse.ArgumentParser()
    parser.add_argument("f_input")
    parser.add_argument("-d", "--debug", dest="debug", action="store_true")
    parser.add_argument("-s", "--stabilise", dest="stab", action="store_true")
    parser.add_argument("-t", "--starttime", dest="starttime")
    args = parser.parse_args()
    if args.debug:
        logging.basicConfig(format='[%(asctime)s]\n%(message)s', level=logging.DEBUG, datefmt='%d/%m/%Y %H:%M:%S')
    else:
        logging.basicConfig(format='[%(asctime)s]\n%(message)s', level=logging.INFO, datefmt='%d/%m/%Y %H:%M:%S')
    file = args.f_input
    infos = get_infos(file)
    # interlaced = is_interlaced(file, infos)
    cropsize = cropping(file, infos)
    volumes = volume_audio(file, infos)
    if args.stab:
        stabilization(file)
    if not args.starttime:
        for track in infos['audio']:
            convert_audio(file, track['index'], volumes[track['index']], track['channels'], track['channel_layout'], track['language'], track['title'])
    if args.starttime:
        vid_part_time = int(args.starttime)
    else:
        vid_part_time = 0
    while vid_part_time < infos['duration']:
        crf = 19
        convert_video(file, infos, vid_part_time, cropsize, crf)
        vid_part_time += 300