Subtitle generation with slightly wrong furigana margin positioning

format.py

@@ -2,15 +2,20 @@ from collections import namedtuple
 # store '{与|あた}えた{使命|しめい}　' as [('与','あた'), ('えた',''), ('使命','しめい'), ('　','')]
 FuriBlock = namedtuple('FuriBlock', ['kanji', 'furi'])
 
+# spb (seconds per beat) is preferred to bpm (beats per minute)
+# spb = 60.0/bpm
 class LyricLine:
 	beat_stamps: list[float] = []  # Start at zero for each line, do real timing via get_timestamps()
 	translated_line: str
+	hiragana_syllables: list[str]  # Allow space entries which will be skipped over when calculating timing
 	romaji_syllables: list[str]  # Allow space entries which will be skipped over when calculating timing
 	furi_blocks: list[FuriBlock]
 
-	def get_timestamps(self, bpm: float, start_offset: float) -> list[float]:
-		spb = 60.0/bpm  # seconds per beat
+	def get_timestamps(self, spb: float, start_offset: float) -> list[float]:
 		return [(spb*beat)+start_offset for beat in self.beat_stamps]
 
+	def get_karaoke_centiseconds(self, spb: float) -> list[float]:
+		return [int(spb*beat*100) for beat in self.beat_stamps]
+
 class LyricTrack:
 	lines: list[LyricLine]

japanese_converters.py

@@ -37,7 +37,7 @@ def parse_jp_text(text: str) -> list[tuple[str, str]]:
 # Our custom word overrides have two levels:
 # - One is a simple search-replace to turn matches into manual furigana "{kanji|furi}" format. This could have false hits on short words.
 # - The latter is to override a word's kana post-tokenization, which requires it to be a dictionary word with multiple readings.
-word_overrides = {'主': 'しゅ'}
+word_overrides = {'私': 'わたし', '主': 'しゅ'}
 re_manual_furi = re.compile(r'{(.+?)\|(.+?)}')
 
 def manual_furi_string_to_blocks(line: str) -> list[FuriBlock]:
@@ -141,10 +141,17 @@ def parse_japanese_line(line: str):
 	for kanji, hiragana in output['word_pairs']:
 		output['furi_blocks'] += word_to_furi_blocks(kanji, hiragana)
 	
-	# Create word-spaced romaji syllables
+	# Create word-spaced hiragana and romaji syllables
+	output['hiragana_syllables'] = []  # Will have spaces mixed in so must be iterated for timing
 	output['romaji_syllables'] = []  # Will have spaces mixed in so must be iterated for timing
 	for _, hiragana in output['word_pairs']:
-		output['romaji_syllables'] += [to_romaji(s) for syl in kana_to_syllable_list(hiragana) if (s:= syl.strip())]
-		if output['romaji_syllables'][-1] != ' ':
+		l = [s for syl in kana_to_syllable_list(hiragana) if (s:= syl.strip())]
+		output['hiragana_syllables'] += l
+		output['romaji_syllables'] += [to_romaji(s) for syl in l if (s:= syl.strip())]
+		if len(l) > 0:
+			output['hiragana_syllables'].append(' ')
 			output['romaji_syllables'].append(' ')
+	if len(output['romaji_syllables']) > 0:  # remove trailing space
+		output['hiragana_syllables'].pop()
+		output['romaji_syllables'].pop()
 	return output

subtitle_generator.py

@@ -34,25 +34,92 @@ format_defaults = {
 	'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)
+	preamble = ass_preamble.format(**format_dict)
 
 	# Kanji Furigana layout stuff
-	size_kanji = format_dict['KanjiSize']
-	size_furi = format_dict['FuriSize']
+	size_kanji_x = format_dict['KanjiSize']  # TODO: work out scaling factor for fullwidth from point size
+	size_furi_x = format_dict['FuriSize']  # TODO: work out scaling factor for fullwidth from point size
+	res_x = format_dict['PlayResX']
 
 	with open(filename, 'w') as file:
 		file.write(preamble)
-		# for line in lines:
-		# 	#
-		# 	for syllable in line:
-		# 		t, kanji, furi, romaji = syllable
+		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}'
 
-example_layout = '''
+			# 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 = 0 if target_middle_x < 0 else int(target_middle_x*1.57)
+					margin_r = 0 if target_middle_x > 0 else int(-target_middle_x*1.57)
+					furi_lines.append(f'{sub_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)
+
+example_layout = r'''
 Dialogue: 0,0:01:08.00,0:01:26.00,Kanji,,,,,,{\k0}{\K100}雨{\K100}や{\K100}雪{\K100}が{\K100}天{\K100}から{\K100}降{\K100}って{\K100}地{\K100}を{\K100}潤{\K100}し　{\K100}芽{\K100}を{\K100}出{\K100}さ{\K100}せ{\K100}る
 Dialogue: 0,0:01:08.00,0:01:26.00,Furigana,,   0,1130,,,{\k0}{\K100}あめ
 Dialogue: 0,0:01:08.00,0:01:26.00,Furigana,,   0, 900,,,{\k200}{\K100}ゆき