From a204811e5a7c7485534cb819785f4ca70596a713 Mon Sep 17 00:00:00 2001
From: Sting <lallegre26@gmail.com>
Date: Sat, 25 Nov 2023 22:35:31 +0100
Subject: [PATCH] Allow choice of language for uppercase words
---
README.md | 2 ++
autokara/autokara.py | 6 +++++-
autokara/autosyl/LyricsAlignment/utils.py | 9 +++------
autokara/autosyl/LyricsAlignment/wrapper.py | 8 ++++----
autokara/autosyl/segment.py | 4 ++--
autokara/default.conf | 4 ++--
6 files changed, 18 insertions(+), 15 deletions(-)
diff --git a/README.md b/README.md
index 4bc86e1..acc798d 100644
--- a/README.md
+++ b/README.md
@@ -42,7 +42,9 @@ To use the custom phonetic mappings for Japanese Romaji and other non-English la
$ autokara-gen-lang
```
+### Windows
+Still working on that...
## Configuration
diff --git a/autokara/autokara.py b/autokara/autokara.py
index 6077478..d981211 100644
--- a/autokara/autokara.py
+++ b/autokara/autokara.py
@@ -20,6 +20,7 @@ def main(opts=None):
parser.add_argument("-o", "--output", help="Write output to specified file. If absent, overwrite source file")
parser.add_argument("-v","--verbose", action="store_true", help="Increased verbosity")
parser.add_argument("-l","--lang", help="Select language to use (default is Japanese Romaji)")
+ parser.add_argument("--uppercase-lang", help="Language to use when interpretinf uppercase words (fefault is English)")
args = parser.parse_args(opts)
@@ -40,9 +41,12 @@ def main(opts=None):
'bdr': config['Segment'].getboolean('bdr'),
'cuda': config['Segment'].getboolean('cuda'),
'syl_delay': config['Segment'].getint('syl_delay'),
- 'uppercase_english': config['Segment'].getboolean('uppercase_as_english')
+ 'uppercase_lang': config['Segment']['uppercase_lang']
}
+ if args.uppercase_lang :
+ segment_config['uppercase_lang'] = args.uppercase_lang
+
if not args.vocals :
print("Extracting audio from video file...")
diff --git a/autokara/autosyl/LyricsAlignment/utils.py b/autokara/autosyl/LyricsAlignment/utils.py
index d5cb782..748a84a 100644
--- a/autokara/autosyl/LyricsAlignment/utils.py
+++ b/autokara/autosyl/LyricsAlignment/utils.py
@@ -117,14 +117,11 @@ def load_lyrics(lyrics_file):
def write_wav(path, audio, sr):
soundfile.write(path, audio.T, sr, "PCM_16")
-def gen_phone_gt(words, raw_lines, language="jp", uppercase_as_english=True):
+def gen_phone_gt(words, raw_lines, language="jp", uppercase_lang="en"):
- print(f"Translating lyrics to phonemes, language chosen : {language:s}, uppercase as English : {str(uppercase_as_english):s}")
+ print(f"Translating lyrics to phonemes, base language : {language:s}, uppercase language : {uppercase_lang:s}")
g2p = G2p_Wrapper(language=language)
- if uppercase_as_english:
- g2p_uppercase = G2p_Wrapper(language="en")
- else:
- g2p_uppercase = g2p
+ g2p_uppercase = G2p_Wrapper(language=uppercase_lang)
regex_uppercase = re.compile('[^a-zà -ÿ]')
diff --git a/autokara/autosyl/LyricsAlignment/wrapper.py b/autokara/autosyl/LyricsAlignment/wrapper.py
index d936e22..d752ac6 100644
--- a/autokara/autosyl/LyricsAlignment/wrapper.py
+++ b/autokara/autosyl/LyricsAlignment/wrapper.py
@@ -11,10 +11,10 @@ from .model import train_audio_transforms, AcousticModel, BoundaryDetection
np.random.seed(7)
-def preprocess_from_file(audio_file, lyrics_file, word_file=None, language="jp", uppercase_as_english=True):
+def preprocess_from_file(audio_file, lyrics_file, word_file=None, language="jp", uppercase_lang="en"):
y, sr = preprocess_audio(audio_file)
- words, lyrics_p, idx_word_p, idx_line_p = preprocess_lyrics(lyrics_file, word_file, language=language, uppercase_as_english=uppercase_as_english)
+ words, lyrics_p, idx_word_p, idx_line_p = preprocess_lyrics(lyrics_file, word_file, language=language, uppercase_lang=uppercase_lang)
return y, words, lyrics_p, idx_word_p, idx_line_p
@@ -144,7 +144,7 @@ def preprocess_audio(audio_file, sr=22050):
return y, curr_sr
-def preprocess_lyrics(lyrics_lines, word_file=None, language="jp", uppercase_as_english=True):
+def preprocess_lyrics(lyrics_lines, word_file=None, language="jp", uppercase_lang="en"):
#from string import ascii_lowercase
#d = {ascii_lowercase[i]: i for i in range(26)}
#d["'"] = 26
@@ -169,7 +169,7 @@ def preprocess_lyrics(lyrics_lines, word_file=None, language="jp", uppercase_as_
else:
words_lines = full_lyrics.split()
- lyrics_p, words_p, idx_word_p, idx_line_p = utils.gen_phone_gt(words_lines, raw_lines, language=(language if language else "jp"), uppercase_as_english=uppercase_as_english)
+ lyrics_p, words_p, idx_word_p, idx_line_p = utils.gen_phone_gt(words_lines, raw_lines, language=(language if language else "jp"), uppercase_lang=uppercase_lang)
return words_lines, lyrics_p, idx_word_p, idx_line_p
diff --git a/autokara/autosyl/segment.py b/autokara/autosyl/segment.py
index 0376de6..029b69f 100644
--- a/autokara/autosyl/segment.py
+++ b/autokara/autosyl/segment.py
@@ -26,7 +26,7 @@ def segment(songfile, config, reference_syls=None,syls_per_line=10, last_syl_dur
checkpoint_folder = f"{str(Path(__file__).parent):s}/LyricsAlignment/checkpoints"
language = language
- uppercase_as_english = config['uppercase_english']
+ uppercase_lang = config['uppercase_lang']
lyrics_lines = [" ".join([syl[1] for syl in line]) for line in reference_syls]
#print(lyrics_lines)
@@ -39,7 +39,7 @@ def segment(songfile, config, reference_syls=None,syls_per_line=10, last_syl_dur
# lyrics_p: phoneme sequence of the target lyrics
# idx_word_p: indices of word start in lyrics_p
# idx_line_p: indices of line start in lyrics_p
- audio, words, lyrics_p, idx_word_p, idx_line_p = preprocess_from_file(audio_file, lyrics_lines, word_file, language, uppercase_as_english=uppercase_as_english)
+ audio, words, lyrics_p, idx_word_p, idx_line_p = preprocess_from_file(audio_file, lyrics_lines, word_file, language, uppercase_lang=uppercase_lang)
if verbose:
print(lyrics_p)
diff --git a/autokara/default.conf b/autokara/default.conf
index 4d48d0e..e654554 100644
--- a/autokara/default.conf
+++ b/autokara/default.conf
@@ -18,5 +18,5 @@ bdr = true
cuda = false
# Default delay applied to detected syls, in centiseconds
syl_delay = -4
-# Whether to treat uppercase words as English. If false, use song language everywhere
-uppercase_as_english = true
\ No newline at end of file
+# Two-letter code for the language to use for uppercase words. Default is "en" for English.
+uppercase_lang = en
\ No newline at end of file
--
GitLab