KaraokeTestudaiBackend/subtitle_generator.py

# Substation Alpha (ASS) generation

# Colour values are &HAABBGGRR, &HBBGGRR, or &HAA.
# Alpha is actually inverted, i.e. transparency - FF is transparent, 00 is opaque.

ass_preamble = '''[Script Info]
ScriptType: v4.00+
WrapStyle: 0
ScaledBorderAndShadow: yes
YCbCr Matrix: TV.709
PlayResX: {PlayResX}
PlayResY: {PlayResY}

[V4+ Styles]
Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding
Style: Default,{LatinFont},72,&H002A0A00,&H000019FF,&H00FFFFFF,&H00000000,0,0,0,0,100,100,0,0,1,2.5,0,8,30,30,25,1
Style: Kanji,{JapaneseFont},{KanjiSize},&H{KaraokeColourPast},&H{KaraokeColourFuture},&H00FFFFFF,&H00000000,0,0,0,0,100,100,0,0,1,4.0,0,2,30,30,{KanjiVMargin},1
Style: Furigana,{JapaneseFont},{FuriSize},&H{KaraokeColourPast},&H{KaraokeColourFuture},&H00FFFFFF,&H00000000,0,0,0,0,100,100,0,0,1,2.5,0,2,0,0,{FuriVMargin},1
Style: Romaji,{LatinFont},{RomajiSize},&H{KaraokeColourPast},&H{KaraokeColourFuture},&H00FFFFFF,&H00000000,0,0,0,0,100,100,0,0,1,2.5,0,8,30,30,20,1
Style: Translation,{LatinFont},{TranslationSize},&H00FFFFFF,&H000019FF,&H00000000,&H00000000,0,1,0,0,100,100,0,0,1,1.0,3,8,30,30,20,1

[Events]
Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text
'''

format_defaults = {
	'PlayResX': 1280,
	'PlayResY': 720,
	'LatinFont': 'Droid Sans',
	'JapaneseFont': 'Droid Sans Japanese',
	'TranslationSize': 36,
	'RomajiSize': 48,
	'KanjiSize': 72,
	'KanjiVMargin': 20,
	'FuriSize': 36,
	'FuriVMargin': 85,
	'KaraokeColourFuture': '000019FF',
	'KaraokeColourPast': 'E02A0A00',
}

def seconds_to_timestamp(t: float) -> str:
	minutes, seconds = divmod(t, 60)
	hours, minutes = divmod(minutes, 60)
	return f'{int(hours):02}:{int(minutes):02}:{seconds:05.2f}'

from format import LyricTrack
from japanese_converters import kana_to_syllable_list
def generate_ass(filename: str, lyric_track: LyricTrack, format_overloads: dict = None):
	format_dict = format_defaults.copy()
	if format_overloads:
		format_dict.update(format_overloads)
	preamble = ass_preamble.format(**format_dict)

	# Kanji Furigana layout stuff
	pt_to_fullwidth = 55.0/72.0  # This seems right for DroidSansJapanese, probably different for each font
	size_kanji_x = format_dict['KanjiSize'] * pt_to_fullwidth
	size_furi_x = format_dict['FuriSize'] * pt_to_fullwidth
	res_x = format_dict['PlayResX']
	res_xh = res_x/2

	with open(filename, 'w') as file:
		file.write(preamble)
		t = 68.0  # placeholder
		for line in lyric_track.lines:
			t0 = seconds_to_timestamp(t)
			timestamps = line.get_timestamps(0.5, t)
			centiseconds = line.get_karaoke_centiseconds(0.5)
			t = timestamps[-1] + 1.0  # placeholder
			t1 = seconds_to_timestamp(t)
			sub_preamble = f'Dialogue: 0,{t0},{t1}'
			furi_preamble = f'Dialogue: 1,{t0},{t1}'  # Different layer to avoid kanji collision detection

			# Translation line is easy and static
			file.write(f'{sub_preamble},Translation,,,,,,{line.translated_line}\n')

			# Romaji line is also easy, just intersperse durations
			romaji_line = f'{{\\k{centiseconds[0]}}}'
			i = 0  # syllable counter
			for syl in line.romaji_syllables:
				if not syl.strip():
					romaji_line += f'{{\\k0}}{syl}'
					continue
				romaji_line += f'{{\\K{centiseconds[i+1]-centiseconds[i]}}}{syl}'
				i += 1
			file.write(f'{sub_preamble},Romaji,,,,,,{romaji_line}\n')

			# Now for the kanji and furi lines...
			kanji_plain_str = ''.join([b.kanji for b in line.furi_blocks])
			full_kanji_width = len(kanji_plain_str) * size_kanji_x
			kanji_line = f'{{\\k{centiseconds[0]}}}'
			kanji_line_progress = 0  # increment as we go, to track furi position
			furi_lines = []
			i = 0  # syllable counter
			for furi_block in line.furi_blocks:
				if len(furi_block.furi) == 0:  # kana or punctuation, nice and simple!
					syls = kana_to_syllable_list(furi_block.kanji)
					for syl in syls:
						if len(syl.strip()) == 0:  # don't time spaces
							kanji_line += f'{{\\k0}}{syl}'
							kanji_line_progress += len(syl)
						else:
							kanji_line += f'{{\\K{centiseconds[i+1]-centiseconds[i]}}}{syl}'
							kanji_line_progress += len(syl)
							i += 1
				else:  # Kanji block
					i0 = i  # Store this to later calculate block time for the kanji
					syls = kana_to_syllable_list(furi_block.furi)
					furi_line = f'{{\\k{centiseconds[i]}}}'
					furi_chars = 0
					for syl in syls:
						furi_line += f'{{\\K{centiseconds[i+1]-centiseconds[i]}}}{syl}'
						furi_chars += len(syl)
						i += 1
					# Need to calculate kanji block position and span to typeset the furigana above it
					k = furi_block.kanji
					k_start = kanji_line_progress
					kanji_line_progress += len(k)
					k_end = kanji_line_progress
					target_middle_x = (size_kanji_x * (k_end+k_start)/2) - (full_kanji_width/2)  # x=0 at center
					furi_width = furi_chars * size_furi_x
					margin_l = int(res_xh+target_middle_x)
					margin_r = int(res_xh-target_middle_x)
					furi_lines.append(f'{furi_preamble},Furigana,,{margin_l},{margin_r},,,{furi_line}\n')
					kanji_line += f'{{\\K{centiseconds[i]-centiseconds[i0]}}}{k}'
			file.write(f'{sub_preamble},Kanji,,,,,,{kanji_line}\n')
			for line in furi_lines:
				file.write(line)