o i0ddlZddlZddlZddlmZddlmZmZmZm Z ddl Z ddl m Z eGddZ dde jd eed ed eefd Zded fde jdeed eded e ee jeeeefff dZGddZejdZGddZdS)N) dataclass)DictListOptionalTuple)get_assets_pathczeZdZUdZdZeed<dZeed<dZe ed<edZ eed <d Z e ed <d Z e ed <dS) VadOptionsarVAD options. Attributes: threshold: Speech threshold. Silero VAD outputs speech probabilities for each audio chunk, probabilities ABOVE this value are considered as SPEECH. It is better to tune this parameter for each dataset separately, but "lazy" 0.5 is pretty good for most datasets. neg_threshold: Silence threshold for determining the end of speech. If a probability is lower than neg_threshold, it is always considered silence. Values higher than neg_threshold are only considered speech if the previous sample was classified as speech; otherwise, they are treated as silence. This parameter helps refine the detection of speech transitions, ensuring smoother segment boundaries. min_speech_duration_ms: Final speech chunks shorter min_speech_duration_ms are thrown out. max_speech_duration_s: Maximum duration of speech chunks in seconds. Chunks longer than max_speech_duration_s will be split at the timestamp of the last silence that lasts more than 100ms (if any), to prevent aggressive cutting. Otherwise, they will be split aggressively just before max_speech_duration_s. min_silence_duration_ms: In the end of each speech chunk wait for min_silence_duration_ms before separating it speech_pad_ms: Final speech chunks are padded by speech_pad_ms each side g? thresholdN neg_thresholdrmin_speech_duration_msinfmax_speech_duration_simin_silence_duration_msi speech_pad_ms) __name__ __module__ __qualname____doc__r float__annotations__r r intrrrR/var/www/html/auto_sub_bot/venv/lib/python3.11/site-packages/faster_whisper/vad.pyr r s*IuM5"#C####(5<<5///#'S'''M3rr >audio vad_options sampling_ratereturnc | tdi|}|j}|j}|j}|j}|j}d} |j} ||zdz } || zdz } ||z| z d| zz } ||zdz }|dzdz }t|}t}tj |d| |j d| zz f}||}d}g}i}|t|dz d }d}dx}}t|D]\}}||kr|r d}||kr| |z}||kr |s d }| |z|d <.|rm| |z|d z | kr[|r1||d <||i}||krd}n||d <dx}x}}n(| |z|d <||i}dx}x}}d}||kr_|r]|s| |z}| |z|z |kr|}| |z|z |kr||d <|d |d z | kr||i}dx}x}}d}|r)||d z | kr||d <||t|D]\\}}|dkr)ttd|d | z |d <|t|d z kr||d zd |d z }|d| zkr_|d xxt|dzz cc<ttd||d zd |dzz ||d zd <tt!||d | z|d <ttd||d zd | z ||d zd <3tt!||d | z|d <^|S)aThis method is used for splitting long audios into speech chunks using silero VAD. Args: audio: One dimensional float array. vad_options: Options for VAD processing. sampling rate: Sampling rate of the audio. kwargs: VAD options passed as keyword arguments for backward compatibility. Returns: List of dicts containing begin and end samples of each speech chunk. NibrFg333333?g{Gz?Tstartendr)r r r r rrrlen get_vad_modelnppadshapemax enumerateappendrmin)rrrkwargsr r r rrwindow_size_samplesrmin_speech_samplesspeech_pad_samplesmax_speech_samplesmin_silence_samples!min_silence_samples_at_max_speechaudio_length_samplesmodel padded_audio speech_probs triggeredspeechescurrent_speechtemp_endprev_end next_starti speech_probspeechsilence_durations rget_speech_timestampsrF-s" **6** %I-M(?'=)A-M&)??$F&6=--     ! (*AADH(5(:T(A%u:: OOE6 &Q:M)MMNL5&&LIHNI,d33 HHz#L1122; 9 $ $( $HH$$014 9 $ $i $I&9A&=N7 #   $q(N7,CCFXXX (0u%///!#(( %II.8N7+3444:(;a(?u%///!#3444:!  - ' 'Y ' 3.2#a'836WWW##a'836III(0u%"5)N7,CC&''OON333!#3444:!  ( !N7$; ;?Q Q Q 4u'''x(( 6 66!#a;M)M"N"NOOF7O H ! ! !'Aw7&-G !&8"888u %5%:!;!;; +.8AE?736F!6KKLL,,Q((!$,fUm>P.PQQ!!u ,/8AE?736HHII,,Q(( (&-:L*LMMF5MM Orrchunks max_durationc|s*ddgd}tjgtjg|gfSg}g}g}d}d} tjgtj} |D]} || dz| dz ||zkre|| | |z ||z |d}| |z } ||g}|| d| d} | d| dz }|| tj| || d| df} || d| dz z }|| | |z ||z |d}||||fS)zPThis function merges the chunks of audio into chunks of max_duration (s) length.r)offsetdurationsegmentsdtyper&r%)r*arrayfloat32r/ concatenate) rrGrrHchunk_metadata audio_chunkschunks_metadatacurrent_segmentscurrent_durationtotal_duration current_audiochunks rcollect_chunksrZs B   2:.../.1AAALONHRrz222M>> uU| +eGn <]* + +    . . .(=8,}<,N . .N  " "> 2 2 2! !%.5<"?@M$U|eGn<    # #E * * *NeGnuU|&C DEM e uW~ = =   &&&!=0$}4$N >***  ((rc teZdZdZddeededefdZ dd ed e ed e d efd Z dd ed e d efdZ dS)SpeechTimestampsMapz3Helper class to restore original speech timestamps.r#rGrtime_precisionc||_||_g|_g|_d}d}|D]X}||d|z z }|d}|j|d|z |j||z YdS)Nrr%r&)rr]chunk_end_sampletotal_silence_beforer/)selfrGrr] previous_endsilent_samplesrYs r__init__zSpeechTimestampsMap.__init__s*, "$&!  M ME eGn|; ;N rs  !!!!!!............000000  @)-JJ :J*%JJ  $Z JJJJ`%,, 9)9) :9) J9)9) 9)  4 T$sEz"23 34 9)9)9)9)x' ' ' ' ' ' ' ' T    8888888888r