Skip to content

Commit

Permalink
Add files via upload
Browse files Browse the repository at this point in the history
  • Loading branch information
@botbahlul
botbahlul authored May 30, 2023
1 parent f39a99f commit affc9af
Showing 1 changed file with 71 additions and 47 deletions.
118 changes: 71 additions & 47 deletions win/whisper_autosrt.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -27,21 +27,7 @@
from faster_whisper import WhisperModel


VERSION = "0.0.5"

whisper_models = [
"tiny.en",
"tiny",
"base.en",
"base",
"small.en",
"small",
"medium.en",
"medium",
"large-v1",
"large-v2",
"large"
]
VERSION = "0.0.6"

#======================================================== ffmpeg_progress_yield ========================================================#

Expand Down Expand Up @@ -1513,9 +1499,49 @@ def show_error_messages(messages):
print(messages)


def get_video_duration(filename, error_messages_callback=None):
try:
command = f'ffprobe -v error -show_entries format=duration -of default=noprint_wrappers=1:nokey=1 "{filename}"'
result = subprocess.run(command, capture_output=True, text=True, shell=True)

if result.returncode == 0:
duration = float(result.stdout)
return duration
else:
msg = f"Failed to get duration for {filename}."
if error_messages_callback:
error_messages_callback(msg)
return None

except Exception as e:
if error_messages_callback:
error_messages_callback(e)
else:
print(e)
return


def main():
global pbar

whisper_models = [
"tiny.en",
"tiny",
"base.en",
"base",
"small.en",
"small",
"medium.en",
"medium",
"large-v1",
"large-v2",
"large"
]

devices = ["auto", "cuda", "cpu"]

compute_types = ["default", "auto", "int8", "int8_float16", "int16", "float16", "float32"]

if sys.platform == "win32":
stop_ffmpeg_windows(error_messages_callback=show_error_messages)
else:
Expand All @@ -1526,15 +1552,21 @@ def main():

parser = argparse.ArgumentParser()
parser.add_argument('source_path', help="Path to the video or audio files to generate subtitles files (use wildcard for multiple files or separate them with a space character e.g. \"file 1.mp4\" \"file 2.mp4\")", nargs='*')
parser.add_argument('-m', '--model-name', default="small", help="name of the Whisper model to use")
parser.add_argument('-lm', '--list-models', help="List of Whisper models name", action='store_true')
parser.add_argument('-m', '--model-name', default="small", help="name of whisper model to use")
parser.add_argument('-lm', '--list-models', help="List of whisper models name", action='store_true')
parser.add_argument('-d', '--device', default="auto", help="name of the device to use")
parser.add_argument('-ld', '--list-devices', help="List of supported device", action='store_true')
parser.add_argument('-ct', '--compute-type', default="auto", help="name of the compute type (quantization) to use")
parser.add_argument('-lct', '--list-compute-types', help="List of supported compute types", action='store_true')
parser.add_argument('-cpu', '--cpu-threads', default=0, help="Number of threads to use when running on CPU")
parser.add_argument('-nw', '--num-workers', default=1, help="Number of concurrent calls when running the model")
parser.add_argument('-S', '--src-language', help="Language code of the audio language spoken in video/audio source_path", default="en")
parser.add_argument('-D', '--dst-language', help="Desired translation language code for the subtitles", default=None)
parser.add_argument('-lwl', '--list-whisper-languages', help="List all whisper supported languages", action='store_true')
parser.add_argument('-lgl', '--list-google-languages', help="List all google translate supported languages", action='store_true')
parser.add_argument('-lwl', '--list-whisper-languages', help="List all whisper supported languages (src_languages)", action='store_true')
parser.add_argument('-lgl', '--list-google-languages', help="List all google translate supported languages (dst_languages)", action='store_true')
parser.add_argument('-F', '--format', help="Desired subtitle format", default="srt")
parser.add_argument('-lf', '--list-formats', help="List all supported subtitle formats", action='store_true')
parser.add_argument('-C', '--concurrency', help="Number of concurrent API requests to make", type=int, default=10)
parser.add_argument('-C', '--concurrency', help="Number of concurrent translate API requests to make", type=int, default=10)
parser.add_argument('-v', '--version', action='version', version=VERSION)

args = parser.parse_args()
Expand All @@ -1549,16 +1581,26 @@ def main():
elif args.src_language != "auto":
args.src_language = src_language

#model = WhisperModel(model_name, compute_type="float32", cpu_threads=4, num_workers=16)
model = WhisperModel(model_name, compute_type="float32")
model = WhisperModel(model_name, device=args.device, compute_type=args.compute_type, cpu_threads=int(args.cpu_threads), num_workers=int(args.num_workers))

if args.list_models:
print("List of whisper models:")
#for model_name in whisper.available_models():
for model_name in whisper_models:
print(model_name)
return 0

if args.list_devices:
print("List of supported devices:")
for device in devices:
print(device)
return 0

if args.list_compute_types:
print("List of supported compute types:")
for compute_type in compute_types:
print(compute_type)
return 0

whisper_language = WhisperLanguage()
google_language = GoogleLanguage()

Expand Down Expand Up @@ -1651,7 +1693,7 @@ def main():
for not_exist_filepath in not_exist_filepaths:
msg = "{} is not exist".format(not_exist_filepath)
print(msg)
if (not ("*" and "?") in str(args_source_path)):
if (not "*" in str(args_source_path)) and (not "?" in str(args_source_path)):
sys.exit(0)

if not arg_filepaths and not not_exist_filepaths:
Expand All @@ -1673,33 +1715,19 @@ def main():
print("Processing {} :".format(media_filepath))

try:
total_duration = get_video_duration(media_filepath, error_messages_callback=show_error_messages)
segments, info = model.transcribe(media_filepath, beam_size=5, language=src_language, task=task)
#marker='█'
widgets = ["Converting to a temporary WAV file : ", Percentage(), ' ', Bar(marker='#'), ' ', ETA()]
pbar = ProgressBar(widgets=widgets, maxval=100).start()
wav_converter = WavConverter(progress_callback=show_progress, error_messages_callback=show_error_messages)
wav_filepath, sample_rate = wav_converter(media_filepath)
pbar.finish()

reader = wave.open(wav_filepath)
rate = reader.getframerate()
total_duration = reader.getnframes() / rate
reader.close()

segments, info = model.transcribe(wav_filepath, beam_size=5, language=src_language, task=task)

widgets = ["Performing speech recognition : ", Percentage(), ' ', Bar(marker='#'), ' ', ETA()]
pbar = ProgressBar(widgets=widgets, maxval=100).start()

timed_subtitles = []
regions = []
transcripts = []
for segment in segments:
start = segment.start
end = segment.end
text = segment.text
regions.append((start, end))
transcripts.append(text)
progress = int(end)*100/total_duration
regions.append((segment.start, segment.end))
transcripts.append(segment.text)
progress = int(int(segment.end)*100/total_duration)
pbar.update(progress)
pbar.finish()
timed_subtitles = [(r, t) for r, t in zip(regions, transcripts) if t]
Expand All @@ -1712,10 +1740,6 @@ def main():
writer.write(subtitle_filepath)

if do_translate:
# CONCURRENT TRANSLATION USING class SentenceTranslator(object)
# NO NEED TO TRANSLATE ALL transcript IN transcripts
# BECAUSE SOME region IN regions MAY JUST HAVE transcript WITH EMPTY STRING
# JUST TRANSLATE ALREADY CREATED subtitles ENTRIES FROM timed_subtitles
timed_subtitles = writer.timed_subtitles
created_regions = []
created_subtitles = []
Expand Down

0 comments on commit affc9af

Please sign in to comment.