g/dddlZddlZddlZddlmZmZmZddlZde de dej fdZ dde de d ed eed eeef d Z dde de d ee deeee e fe fd ed eed eeefdZdde dee e fdefdZde fdZdZdS)N)OptionalTupleUnionbpayload sampling_ratereturnc |}d}d}dddd|d|d|d d d d g } tj|tjtj 5}||}dddn #1swxYwYn"#t$r}t d|d}~wwxYw|d} t j| t j} | j ddkrt d| S)z? Helper function to read an audio file through ffmpeg. 1f32leffmpeg-izpipe:0-ac-ar-f -hide_banner -loglevelquietpipe:1)stdinstdoutNzFffmpeg was not found but is required to load audio files from filenamera Soundfile is either not in the correct format or is malformed. Ensure that the soundfile has a valid audio file extension (e.g. wav, flac or mp3) and is not corrupted. If reading from a remote URL, ensure that the URL is the full address to **download** the audio file.) subprocessPopenPIPE communicateFileNotFoundError ValueErrornp frombufferfloat32shape) rraracformat_for_conversionffmpeg_commandffmpeg_process output_streamerror out_bytesaudios ^/var/www/html/ai-engine/env/lib/python3.11/site-packages/transformers/pipelines/audio_utils.py ffmpeg_readr+ s`  B B#      N n  nJOJO \ \ \ A`n*66x@@M A A A A A A A A A A A A A A A nnnabbhmmna I M)RZ 0 0E {1~ [   Ls;+A1A% A1%A))A1,A)-A11 B;B  Br chunk_length_sr#ffmpeg_input_deviceffmpeg_additional_argsc#K|}d}|dkrd}n|dkrd}ntd|dtj}|dkrd } |pd } n%|d krd } |pd } n|dkrd} |p t} |gn|}dd| d| d|d|d|ddddddg} | |t t ||z|z} t| | } | D]}|VdS)a Helper function to read audio from a microphone using ffmpeg. The default input device will be used unless another input device is specified using the `ffmpeg_input_device` argument. Uses 'alsa' on Linux, 'avfoundation' on MacOS and 'dshow' on Windows. Arguments: sampling_rate (`int`): The sampling_rate to use when reading the data from the microphone. Try using the model's sampling_rate to avoid resampling later. chunk_length_s (`float` or `int`): The length of the maximum chunk of audio to be sent returned. format_for_conversion (`str`, defaults to `f32le`): The name of the format of the audio samples to be returned by ffmpeg. The standard is `f32le`, `s16le` could also be used. ffmpeg_input_device (`str`, *optional*): The indentifier of the input device to be used by ffmpeg (i.e. ffmpeg's '-i' argument). If unset, the default input device will be used. See `https://www.ffmpeg.org/ffmpeg-devices.html#Input-Devices` for how to specify and list input devices. ffmpeg_additional_args (`list[str]`, *optional*): Additional arguments to pass to ffmpeg, can include arguments like -nostdin for running as a background process. For example, to pass -nostdin to the ffmpeg process, pass in ["-nostdin"]. If passing in flags with multiple arguments, use the following convention (eg ["flag", "arg1", "arg2]). Returns: A generator yielding audio chunks of `chunk_length_s` seconds as `bytes` objects of length `int(round(sampling_rate * chunk_length_s)) * size_of_sample`. r s16ler Unhandled format ` `. Please use `s16le` or `f32le`LinuxalsadefaultDarwin avfoundationz:defaultWindowsdshowNr rr rrz-fflagsnobufferrrrr)rplatformsystem_get_microphone_nameextendintround_ffmpeg_stream)rr,r#r-r.r!r"size_of_sampler>format_input_r$ chunk_leniteratoritems r*ffmpeg_microphonerJ1spD  B B'' ' ) )e.Ceeefff _  F $1  8   $2  9  $>(<(>(>#9#ARRG]        #N(0111E-.899::^KIni88H stream_chunk_sstride_length_sc#K||}n|}t|||||gn|}|dkrtj} d} n(|dkrtj} d} nt d|d||d z }t t ||z| z} t|t tfr||g}t t ||d z| z} t t ||d z| z} tj }tj | }t|| | | fd D]}}tj |d| |d<|dd | z|dd | zf|d<||d<||z }tj |d|zzkry|V~dS)a Helper function to read audio from a microphone using ffmpeg. This will output `partial` overlapping chunks starting from `stream_chunk_s` (if it is defined) until `chunk_length_s` is reached. It will make use of striding to avoid errors on the "sides" of the various chunks. The default input device will be used unless another input device is specified using the `ffmpeg_input_device` argument. Uses 'alsa' on Linux, 'avfoundation' on MacOS and 'dshow' on Windows. Arguments: sampling_rate (`int`): The sampling_rate to use when reading the data from the microphone. Try using the model's sampling_rate to avoid resampling later. chunk_length_s (`float` or `int`): The length of the maximum chunk of audio to be sent returned. This includes the eventual striding. stream_chunk_s (`float` or `int`): The length of the minimal temporary audio to be returned. stride_length_s (`float` or `int` or `(float, float)`, *optional*): The length of the striding to be used. Stride is used to provide context to a model on the (left, right) of an audio sample but without using that part to actually make the prediction. Setting this does not change the length of the chunk. format_for_conversion (`str`, *optional*, defaults to `f32le`): The name of the format of the audio samples to be returned by ffmpeg. The standard is `f32le`, `s16le` could also be used. ffmpeg_input_device (`str`, *optional*): The identifier of the input device to be used by ffmpeg (i.e. ffmpeg's '-i' argument). If unset, the default input device will be used. See `https://www.ffmpeg.org/ffmpeg-devices.html#Input-Devices` for how to specify and list input devices. ffmpeg_additional_args (`list[str]`, *optional*): Additional arguments to pass to ffmpeg, can include arguments like -nostdin for running as a background process. For example, to pass -nostdin to the ffmpeg process, pass in ["-nostdin"]. If passing in flags with multiple arguments, use the following convention (eg ["flag", "arg1", "arg2]). Return: A generator yielding dictionaries of the following form `{"sampling_rate": int, "raw": np.array(), "partial" bool}` With optionally a `"stride" (int, int)` key if `stride_length_s` is defined. `stride` and `raw` are all expressed in `samples`, and `partial` is a boolean saying if the current yield item is a whole chunk, or a partial temporary result to be later replaced by another larger chunk. N)r#r-r.r0r1r r2r3r4r)secondsT)stridestreamraw)dtyperRr )rJrint16rrrArB isinstancefloatdatetimenow timedeltachunk_bytes_iterr)rr,rLrMr#r-r.chunk_s microphonerUrDrG stride_left stride_right audio_timedeltarIs r*ffmpeg_microphone_liverds`!  "3/%;%CrrI_ J'' ' ) ) e.Ceeefff(1,E-.899::^KI/C<00=*O<eMOA,>>??@@>QKu]_Q-??@@AANRL"&&((J  w / / /E Y \?Zcghhh  mDKu===U N1  / N1  / X!._e   " "Z"u*%< < <    rKFrGrRrSc#Kd}|\}}||z|krtd|d|d|d}|D]}||z }|r(t||kr|df}|d||ddV1t||krA||f}|d||d } |rd | d <| V|}|||z |z d}t||kAt||kr||dfd } |rd | d <| VdSdS) z Reads raw bytes from an iterator and does chunks of length `chunk_len`. Optionally adds `stride` to each chunks to get overlaps. `stream` is used to return partial results even if a full `chunk_len` is not yet available. rKz5Stride needs to be strictly smaller than chunk_len: (z, z) vs rNT)rTrRpartial)rTrRFrf)rlen) rHrGrRrSaccr`ra _stride_leftrTrIs r*r]r]s C &K\!Y.. qK q qS_ q qfo q q   LDD s   Dc#hh**"A&Fjyj/VMM M M M Mc((i''& 5":I:&AA,&+DO * )k1L@BBCc((i'' 3xx+|Q&788  $#DO rKbuflenc#Kd} tj|tj|5} |j|}|dkrn|V& ddddS#1swxYwYdS#t $r}t d|d}~wwxYw)zJ Internal function to create the generator of data through ffmpeg i)rbufsizeTrKNzHffmpeg was not found but is required to stream audio files from filename)rrrrreadrr)r$rjrlr%rTr's r*rCrCsGp  nZ_g V V V Zh $+0088#::                      pppcddjoops:!A*(A A*A!!A*$A!%A** B 4BB cNgd} tj|dtjd}d|jD}|r8|ddd}t d |d |Sn#t$rt d YnwxYwd S) z3 Retrieve the microphone name in Windows . )r z -list_devicestruerr;r Tzutf-8)textstderrencodingcg|]}d|v| S)z(audio)).0lines r* z(_get_microphone_name.. s$eeeTS\`dSdSdDSdSdSdrKr"rPzUsing microphone: zaudio=zOffmpeg was not found. Please install it or make sure it is in your system PATH.r7)rrunrrr splitlinessplitprintr)commandffmpeg_devicesmicrophone_linesmicrophone_names r*r?r?sKJJG a#d:?]deeeee^-B-M-M-O-Oeee  ..q177<BB"!B")r NN)NNr NN)F)rZr=rtypingrrrnumpyrbytesrAarrayr+rYstrlistrJrdboolr]rCr?rurKr*rs))))))))))$%$$$$$$T"))-26 RRRRR"# R %T#Y/ RRRRp%)CG!()-26\\\\SM\eE%,$7$>?@ \  \ "# \%T#Y/\\\\~  # uS#X PT    F p3 p p p p rK