dasemu/Transcriptarr
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

add queue tracking for Bazarr

Browse Source 
General clean up, removed the old array that tracked items and added Bazarr.  Bazarr doesn't obey the CONCURRENT_TRANSCRIPTIONS, which is probably best since those have a fixed timeout they need to respond by.
This commit is contained in:
McCloudS
2024-04-20 11:28:26 -06:00
committed by GitHub
parent 2e6bf94b09
commit 8a6e625365
1 changed files with 23 additions and 39 deletions
Download Patch File Download Diff File Expand all files Collapse all files

62
subgen.py
View File

@@ -109,7 +109,6 @@ update_env_variables()
app = FastAPI() app = FastAPI()
model = None model = None
files_to_transcribe = []
in_docker = os.path.exists('/.dockerenv') in_docker = os.path.exists('/.dockerenv')
docker_status = "Docker" if in_docker else "Standalone" docker_status = "Docker" if in_docker else "Standalone"
@@ -122,8 +121,11 @@ task_queue = queue.Queue()
def transcription_worker(): def transcription_worker():
while True: while True:
task = task_queue.get() task = task_queue.get()
gen_subtitles(task['path'], task['transcribe_or_translate'], task['force'],task['force_language']) if 'Bazarr-' in task['path']:
task_queue.task_done() logging.info(f"Skipping processing for {task['path']} as it is handled by ASR.")
else:
gen_subtitles(task['path'], task['transcribe_or_translate'], task['force_language'])
task_queue.task_done()
# show queue # show queue
logging.debug(f"There are {task_queue.qsize()} tasks left in the queue.") logging.debug(f"There are {task_queue.qsize()} tasks left in the queue.")
@@ -309,7 +311,6 @@ def receive_tautulli_webhook(
fullpath = file fullpath = file
logging.debug("Path of file: " + fullpath) logging.debug("Path of file: " + fullpath)
# gen_subtitles(path_mapping(fullpath), transcribe_or_translate, True)
gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True) gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True)
else: else:
return { return {
@@ -337,7 +338,6 @@ def receive_plex_webhook(
fullpath = get_plex_file_name(plex_json['Metadata']['ratingKey'], plexserver, plextoken) fullpath = get_plex_file_name(plex_json['Metadata']['ratingKey'], plexserver, plextoken)
logging.debug("Path of file: " + fullpath) logging.debug("Path of file: " + fullpath)
# gen_subtitles(path_mapping(fullpath), transcribe_or_translate, True)
gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True) gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True)
refresh_plex_metadata(plex_json['Metadata']['ratingKey'], plexserver, plextoken) refresh_plex_metadata(plex_json['Metadata']['ratingKey'], plexserver, plextoken)
logging.info(f"Metadata for item {plex_json['Metadata']['ratingKey']} refreshed successfully.") logging.info(f"Metadata for item {plex_json['Metadata']['ratingKey']} refreshed successfully.")
@@ -363,7 +363,6 @@ def receive_jellyfin_webhook(
fullpath = get_jellyfin_file_name(ItemId, jellyfinserver, jellyfintoken) fullpath = get_jellyfin_file_name(ItemId, jellyfinserver, jellyfintoken)
logging.debug(f"Path of file: {fullpath}") logging.debug(f"Path of file: {fullpath}")
# gen_subtitles(path_mapping(fullpath), transcribe_or_translate, True)
gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True) gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True)
try: try:
refresh_jellyfin_metadata(ItemId, jellyfinserver, jellyfintoken) refresh_jellyfin_metadata(ItemId, jellyfinserver, jellyfintoken)
@@ -397,7 +396,6 @@ def receive_emby_webhook(
if event == "library.new" and procaddedmedia or event == "playback.start" and procmediaonplay: if event == "library.new" and procaddedmedia or event == "playback.start" and procmediaonplay:
logging.debug("Path of file: " + fullpath) logging.debug("Path of file: " + fullpath)
# gen_subtitles(path_mapping(fullpath), transcribe_or_translate, True)
gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True) gen_subtitles_queue(path_mapping(fullpath), transcribe_or_translate, True)
return "" return ""
@@ -424,14 +422,17 @@ def asr(
try: try:
logging.info(f"Transcribing file from Bazarr/ASR webhook") logging.info(f"Transcribing file from Bazarr/ASR webhook")
result = None result = None
random_name = random.choices("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890", k=6) random_name = ''.join(random.choices("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890", k=6))
if force_detected_language_to: if force_detected_language_to:
language = force_detected_language_to language = force_detected_language_to
start_time = time.time() start_time = time.time()
start_model() start_model()
files_to_transcribe.insert(0, f"Bazarr-asr-{random_name}")
task_id = { 'path': f"Bazarr-asr-{random_name}" }
task_queue.put(task_id)
audio_data = np.frombuffer(audio_file.file.read(), np.int16).flatten().astype(np.float32) / 32768.0 audio_data = np.frombuffer(audio_file.file.read(), np.int16).flatten().astype(np.float32) / 32768.0
if model_prompt: if model_prompt:
custom_prompt = greetings_translations.get(language, '') or custom_model_prompt custom_prompt = greetings_translations.get(language, '') or custom_model_prompt
@@ -446,8 +447,7 @@ def asr(
except Exception as e: except Exception as e:
logging.info(f"Error processing or transcribing Bazarr {audio_file.filename}: {e}") logging.info(f"Error processing or transcribing Bazarr {audio_file.filename}: {e}")
finally: finally:
if f"Bazarr-asr-{random_name}" in files_to_transcribe: task_queue.task_done()
files_to_transcribe.remove(f"Bazarr-asr-{random_name}")
delete_model() delete_model()
if result: if result:
return StreamingResponse( return StreamingResponse(
@@ -470,8 +470,11 @@ def detect_language(
logging.info(f"Detect language is set to detect on the first {detect_language_length} seconds of the audio.") logging.info(f"Detect language is set to detect on the first {detect_language_length} seconds of the audio.")
try: try:
start_model() start_model()
random_name = random.choices("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890", k=6) random_name = ''.join(random.choices("abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ1234567890", k=6))
files_to_transcribe.insert(0, f"Bazarr-detect-language-{random_name}")
task_id = { 'path': f"Bazarr-detect-language-{random_name}" }
task_queue.put(task_id)
audio_data = np.frombuffer(audio_file.file.read(), np.int16).flatten().astype(np.float32) / 32768.0 audio_data = np.frombuffer(audio_file.file.read(), np.int16).flatten().astype(np.float32) / 32768.0
detected_lang_code = model.transcribe_stable(whisper.pad_or_trim(audio_data, detect_language_length * 16000), input_sr=16000).language detected_lang_code = model.transcribe_stable(whisper.pad_or_trim(audio_data, detect_language_length * 16000), input_sr=16000).language
@@ -479,8 +482,7 @@ def detect_language(
logging.info(f"Error processing or transcribing Bazarr {audio_file.filename}: {e}") logging.info(f"Error processing or transcribing Bazarr {audio_file.filename}: {e}")
finally: finally:
if f"Bazarr-detect-language-{random_name}" in files_to_transcribe: task_queue.task_done()
files_to_transcribe.remove(f"Bazarr-detect-language-{random_name}")
delete_model() delete_model()
return {"detected_language": whisper_languages.get(detected_lang_code, detected_lang_code) , "language_code": detected_lang_code} return {"detected_language": whisper_languages.get(detected_lang_code, detected_lang_code) , "language_code": detected_lang_code}
@@ -492,7 +494,7 @@ def start_model():
model = stable_whisper.load_faster_whisper(whisper_model, download_root=model_location, device=transcribe_device, cpu_threads=whisper_threads, num_workers=concurrent_transcriptions, compute_type=compute_type) model = stable_whisper.load_faster_whisper(whisper_model, download_root=model_location, device=transcribe_device, cpu_threads=whisper_threads, num_workers=concurrent_transcriptions, compute_type=compute_type)
def delete_model(): def delete_model():
if clear_vram_on_complete and len(files_to_transcribe) == 0: if clear_vram_on_complete and task_queue.qsize() == 0:
global model global model
logging.debug("Queue is empty, clearing/releasing VRAM") logging.debug("Queue is empty, clearing/releasing VRAM")
model = None model = None
@@ -509,25 +511,17 @@ def write_lrc(result, file_path):
fraction = int((segment.start - int(segment.start)) * 100) fraction = int((segment.start - int(segment.start)) * 100)
file.write(f"[{minutes:02d}:{seconds:02d}.{fraction:02d}] {segment.text}\n") file.write(f"[{minutes:02d}:{seconds:02d}.{fraction:02d}] {segment.text}\n")
def gen_subtitles(file_path: str, transcription_type: str, add_to_front=True, force_language=None) -> None: def gen_subtitles(file_path: str, transcription_type: str, force_language=None) -> None:
"""Generates subtitles for a video file. """Generates subtitles for a video file.
Args: Args:
file_path: str - The path to the video file. file_path: str - The path to the video file.
transcription_type: str - The type of transcription or translation to perform. transcription_type: str - The type of transcription or translation to perform.
add_to_front: bool - Whether to add the file to the front of the transcription queue. Default is True.
force_language: str - The language to force for transcription or translation. Default is None. force_language: str - The language to force for transcription or translation. Default is None.
""" """
try: try:
if add_to_front:
files_to_transcribe.insert(0, file_path)
else:
files_to_transcribe.append(file_path)
logging.info(f"Added {os.path.basename(file_path)} for transcription.") logging.info(f"Added {os.path.basename(file_path)} for transcription.")
#logging.info(f"{len(files_to_transcribe)} files in the queue for transcription")
logging.info(f"Transcribing file: {os.path.basename(file_path)}") logging.info(f"Transcribing file: {os.path.basename(file_path)}")
start_time = time.time() start_time = time.time()
@@ -562,21 +556,15 @@ def gen_subtitles(file_path: str, transcription_type: str, add_to_front=True, fo
logging.info(f"Error processing or transcribing {file_path}: {e}") logging.info(f"Error processing or transcribing {file_path}: {e}")
finally: finally:
if file_path in files_to_transcribe:
files_to_transcribe.remove(file_path)
delete_model() delete_model()
def gen_subtitles_queue(file_path: str, transcription_type: str, add_to_front=True, force_language=None) -> None: def gen_subtitles_queue(file_path: str, transcription_type: str, force_language=None) -> None:
global task_queue global task_queue
if not has_audio(file_path): if not has_audio(file_path):
logging.debug(f"{file_path} doesn't have any audio to transcribe!") logging.debug(f"{file_path} doesn't have any audio to transcribe!")
return return
if file_path in files_to_transcribe:
logging.info(f"File {os.path.basename(file_path)} is already in the transcription list. Skipping.")
return
message = None message = None
if has_subtitle_language(file_path, skipifinternalsublang): if has_subtitle_language(file_path, skipifinternalsublang):
message = f"{file_path} already has an internal subtitle we want, skipping generation" message = f"{file_path} already has an internal subtitle we want, skipping generation"
@@ -591,7 +579,6 @@ def gen_subtitles_queue(file_path: str, transcription_type: str, add_to_front=Tr
task = { task = {
'path': file_path, 'path': file_path,
'transcribe_or_translate': transcription_type, 'transcribe_or_translate': transcription_type,
'force': add_to_front,
'force_language':force_language 'force_language':force_language
} }
task_queue.put(task) task_queue.put(task)
@@ -766,8 +753,7 @@ if monitor:
if has_audio(file_path): if has_audio(file_path):
# Call the gen_subtitles function # Call the gen_subtitles function
logging.info(f"File: {path_mapping(file_path)} was added") logging.info(f"File: {path_mapping(file_path)} was added")
# gen_subtitles(path_mapping(file_path), transcribe_or_translate, False) gen_subtitles_queue(path_mapping(file_path), transcribe_or_translate)
gen_subtitles_queue(path_mapping(file_path), transcribe_or_translate, False)
def on_created(self, event): def on_created(self, event):
self.create_subtitle(event) self.create_subtitle(event)
def on_modified(self, event): def on_modified(self, event):
@@ -782,13 +768,11 @@ def transcribe_existing(transcribe_folders, forceLanguage=None):
for root, dirs, files in os.walk(path): for root, dirs, files in os.walk(path):
for file in files: for file in files:
file_path = os.path.join(root, file) file_path = os.path.join(root, file)
# gen_subtitles(path_mapping(file_path), transcribe_or_translate, False, forceLanguage) gen_subtitles_queue(path_mapping(file_path), transcribe_or_translate, forceLanguage)
gen_subtitles_queue(path_mapping(file_path), transcribe_or_translate, False, forceLanguage)
# if the path specified was actually a single file and not a folder, process it # if the path specified was actually a single file and not a folder, process it
if os.path.isfile(path): if os.path.isfile(path):
if has_audio(path): if has_audio(path):
# gen_subtitles(path_mapping(path), transcribe_or_translate, False, forceLanguage) gen_subtitles_queue(path_mapping(path), transcribe_or_translate, forceLanguage)
gen_subtitles_queue(path_mapping(path), transcribe_or_translate, False, forceLanguage)
# Set up the observer to watch for new files # Set up the observer to watch for new files
if monitor: if monitor:
observer = Observer() observer = Observer()