Skip to content

Commit

Permalink
Fixed youtube title to artist and song guessing for convenience
Browse files Browse the repository at this point in the history
  • Loading branch information
@beveradb
beveradb committed Jul 6, 2023
1 parent 77a5965 commit ed59c72
Show file tree
Hide file tree
Showing 4 changed files with 276 additions and 144 deletions.
51 changes: 34 additions & 17 deletions karaoke_generator/generator.py
View file
Original file line number Diff line number Diff line change
@@ -1,5 +1,8 @@
import os
import sys
import json
import re
import regex
import urllib
import shutil
import logging
Expand All @@ -18,6 +21,7 @@ def __init__(
song_artist=None,
song_title=None,
genius_api_token=None,
spotify_cookie=None,
model_name="UVR_MDXNET_KARA_2",
model_file_dir="/tmp/audio-separator-models",
cache_dir="/tmp/karaoke-generator-cache",
Expand All @@ -40,7 +44,9 @@ def __init__(
self.cache_dir = cache_dir
self.output_dir = output_dir

self.genius_api_token = genius_api_token
self.genius_api_token = os.getenv("GENIUS_API_TOKEN", default=genius_api_token)
self.spotify_cookie = os.getenv("SPOTIFY_COOKIE_SP_DC", default=spotify_cookie)

self.song_artist = song_artist
self.song_title = song_title

Expand Down Expand Up @@ -88,6 +94,7 @@ def transcribe_lyrics(self):
transcriber = LyricsTranscriber(
self.audio_file,
genius_api_token=self.genius_api_token,
spotify_cookie=self.spotify_cookie,
song_artist=self.song_artist,
song_title=self.song_title,
output_dir=self.output_dir,
Expand All @@ -114,10 +121,7 @@ def transcribe_lyrics(self):
self.logger.debug(f"Whisper transcription output JSON file: {transcription_metadata['whisper_json_filepath']}")
self.logger.debug(f"MidiCo LRC output file: {transcription_metadata['midico_lrc_filepath']}")
self.logger.debug(f"Genius lyrics output file: {transcription_metadata['genius_lyrics_filepath']}")

if self.output_dir:
shutil.copy(self.output_values["transcription_metadata"]["midico_lrc_filepath"], self.output_dir)
shutil.copy(self.output_values["transcription_metadata"]["genius_lyrics_filepath"], self.output_dir)
self.logger.debug(f"Spotify lyrics output file: {transcription_metadata['genius_lyrics_filepath']}")

def separate_audio(self):
if self.audio_file is None or not os.path.isfile(self.audio_file):
Expand Down Expand Up @@ -153,9 +157,9 @@ def download_youtube_video(self):
if os.path.isfile(ydl_info_cache_file):
self.logger.debug(f"Reading ydl info from cache: {ydl_info_cache_file}")
with open(ydl_info_cache_file, "r") as cache_file:
info = json.load(cache_file)
self.youtube_video_file = info["download_filepath"]
self.output_filename_slug = info["output_filename_slug"]
youtube_info = json.load(cache_file)
self.youtube_video_file = youtube_info["download_filepath"]
self.output_filename_slug = youtube_info["output_filename_slug"]
else:
self.logger.debug(f"No existing YDL info file found")

Expand All @@ -167,38 +171,51 @@ def download_youtube_video(self):
ydl_opts = {
"logger": YoutubeDLLogger(self.logger),
"format": "best",
"parse-metadata": "pre_process",
"outtmpl": os.path.join(self.cache_dir, "%(id)s-%(title)s.%(ext)s"),
}

# Download the original highest quality file
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
info = ydl.extract_info(self.youtube_url, download=False)
youtube_info = ydl.extract_info(self.youtube_url, download=False)

temp_download_filepath = ydl.prepare_filename(info)
temp_download_filepath = ydl.prepare_filename(youtube_info)
self.logger.debug(f"temp_download_filepath: {temp_download_filepath}")

download_file_extension = os.path.splitext(os.path.basename(temp_download_filepath))[1]
self.logger.debug(f"download_file_extension: {download_file_extension}")

# Create a slugified filename prefix to get rid of spaces and unicode characters from youtube titles
info["output_filename_slug"] = info["id"] + "-" + slugify.slugify(info["title"], lowercase=False)
self.output_filename_slug = info["output_filename_slug"]
youtube_info["output_filename_slug"] = youtube_info["id"] + "-" + slugify.slugify(youtube_info["title"], lowercase=False)
self.output_filename_slug = youtube_info["output_filename_slug"]
self.logger.debug(f"output_filename_slug: {self.output_filename_slug}")

# but retain original file extension (which may vary depending on the video format youtube returns)
info["download_filepath"] = os.path.join(self.cache_dir, self.output_filename_slug + download_file_extension)
self.logger.debug(f"download_filepath: {info['download_filepath'] }")
youtube_info["download_filepath"] = os.path.join(self.cache_dir, self.output_filename_slug + download_file_extension)
self.logger.debug(f"download_filepath: {youtube_info['download_filepath'] }")

# Save the ydl.extract_info output to cache file
self.logger.debug(f"Saving sanitized YT-DLP info to cache: {ydl_info_cache_file}")
with open(ydl_info_cache_file, "w") as cache_file:
json.dump(ydl.sanitize_info(info), cache_file, indent=4)
json.dump(ydl.sanitize_info(youtube_info), cache_file, indent=4)

ydl.download([self.youtube_url])
shutil.move(temp_download_filepath, info["download_filepath"])
self.youtube_video_file = info["download_filepath"]
shutil.move(temp_download_filepath, youtube_info["download_filepath"])
self.youtube_video_file = youtube_info["download_filepath"]
self.logger.debug(f"successfully downloaded youtube video to path: {self.youtube_video_file}")

if self.song_title is None:
self.logger.debug(f"Song title not specified, attempting to split from YouTube title: {youtube_info['title']}")
# Define the hyphen variations pattern
hyphen_pattern = regex.compile(r" [^[:ascii:]-_\p{Dash}] ")
# Split the string using the hyphen variations pattern
title_parts = hyphen_pattern.split(youtube_info["title"])

self.song_artist = title_parts[0]
self.song_title = title_parts[1]

print(f"Guessed metadata from title: Artist: {self.song_artist}, Title: {self.song_title}")

# Extract audio to WAV file using ffmpeg
self.audio_file = os.path.join(self.cache_dir, self.output_filename_slug + ".wav")
if os.path.isfile(self.audio_file):
Expand Down
23 changes: 12 additions & 11 deletions karaoke_generator/utils/cli.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -32,19 +32,24 @@ def main():
parser.add_argument(
"--song_artist",
default=None,
help="Optional: specify song artist for Genius lyrics lookup and auto-correction",
help="Optional: specify song artist for lyrics lookup and auto-correction",
)
parser.add_argument(
"--song_title",
default=None,
help="Optional: specify song title for Genius lyrics lookup and auto-correction",
help="Optional: specify song title for lyrics lookup and auto-correction",
)

parser.add_argument(
"--genius_api_token",
default=None,
help="Optional: specify Genius API token for lyrics lookup and auto-correction",
)
parser.add_argument(
"--spotify_cookie",
default=None,
help="Optional: specify Spotify SP_DC cookie value for lyrics lookup and auto-correction",
)

parser.add_argument(
"--model_name",
Expand Down Expand Up @@ -77,21 +82,17 @@ def main():
song_artist=args.song_artist,
song_title=args.song_title,
genius_api_token=args.genius_api_token,
spotify_cookie=args.spotify_cookie,
model_name=args.model_name,
model_file_dir=args.model_file_dir,
cache_dir=args.cache_dir,
output_dir=args.output_dir,
)
output_files = generator.generate()

logger.info(f"Karaoke generation complete! Outputs: ")
for key in output_files:
if type(output_files[key]) is dict:
for key2 in output_files[key]:
print(f"{key} / {key2}: {output_files[key][key2]}")
else:
print(f"{key}: {output_files[key]}")
outputs = generator.generate()

logger.info(f"Karaoke generation complete!")

logger.debug(f"Output folder: {outputs['output_dir']}")

if __name__ == "__main__":
main()
Loading

0 comments on commit ed59c72

Please sign in to comment.