qifdZddlZddlmZddlZddlmZddlm Z m Z dZ ddl Z Gd d ZGd d e Zy#e $r7Zejdeejded edZ[wwxYw)zSilero Voice Activity Detection (VAD) implementation for Pipecat. This module provides a VAD analyzer based on the Silero VAD ONNX model, which can detect voice activity in audio streams with high accuracy. Supports 8kHz and 16kHz sample rates. N)Optional)logger) VADAnalyzer VADParamsg@z Exception: zAIn order to use Silero VAD, you need to `pip install pipecat-ai`.zMissing module(s): c8eZdZdZddZdefdZd dZdefdZy) SileroOnnxModelzONNX runtime wrapper for the Silero VAD model. Provides voice activity detection using the pre-trained Silero VAD model with ONNX runtime for efficient inference. Handles model state management and input validation for audio processing. c tj}d|_d|_|r5dtjvrtj |dg||_ntj |||_|jddg|_y)zInitialize the Silero ONNX model. Args: path: Path to the ONNX model file. force_onnx_cpu: Whether to force CPU execution provider. CPUExecutionProvider) providers sess_options)r @>N) onnxruntimeSessionOptionsinter_op_num_threadsintra_op_num_threadsget_available_providersInferenceSessionsession reset_states sample_rates)selfpathforce_onnx_cpuoptss J/opt/pipecat/venv/lib/python3.12/site-packages/pipecat/audio/vad/silero.py__init__zSileroOnnxModel.__init__*s))+$%!$%! 4 8[8[8]]&77!7 8tDL'774PDL !5Msrcntj|dk(rtj|d}tj|dkDrtd|j ||j vrtd|j d|tj |dz dkDr td||fS) z)Validate and preprocess input audio data.r rz*Too many dimensions for input audio chunk zSupported sampling rates: z (or multiple of 16000)g@?@zInput audio chunk is too short)npndim expand_dims ValueErrordimrshape)rxr s r_validate_inputzSileroOnnxModel._validate_input?s 771:?q!$A 771:>I!%%'ST T T&& &,T->->,??VW   A  &=> >"u rctjd|dfd|_tj|dfd|_d|_d|_y)zReset the internal model states. Args: batch_size: Batch size for state initialization. Defaults to 1. r"float32dtyperN)r#zeros_state_context_last_sr_last_batch_size)r batch_sizes rrzSileroOnnxModel.reset_statesOsB hh:s39E *a B  !rc|j||\}}|dk(rdnd}tj|d|k7r%tdtj|ddtj|d}|dk(rdnd }|js|j ||j r |j |k7r|j ||jr |j|k7r|j |tj|jd stj||fd |_tj|j|fd }|dvrN||jtj|d d}|jjd|}|\}} | |_ n t|d| df|_||_||_|S)z*Process audio input through the VAD model.rzProvided number of samples is z< (Supported values: 256 for 8000 sample rate, 512 for 16000)r@ r r-r.)axis)rrint64)inputstater N.)r*r#r(r&r4rr3r2r0 concatenater1arrayrrun) rr)r num_samplesr5 context_size ort_inputsort_outsoutr?s r__call__zSileroOnnxModel.__call__Zs$$Q+25[cc 88A;r?k )0!R0AA}~ XXa[^ 5[rb $$   j ) MM  3   j )  ! !(=(=(K   j )xx &q)HHj,%?yQDM NNDMM1-A 6  #$t{{"((2U\B]^J||''j9H!JCDK, # }~-.  * rN)T)r ) __name__ __module__ __qualname____doc__rintr*rrHrrrr"s)**S "%c%rrcheZdZdZddddeedeeffdZdeffd Zdefd Z de fd Z xZ S) SileroVADAnalyzeraVoice Activity Detection analyzer using the Silero VAD model. Implements VAD analysis using the pre-trained Silero ONNX model for accurate voice activity detection. Supports 8kHz and 16kHz sample rates with automatic model state management and periodic resets. N sample_rateparamsrRrSct|||tjdd}d} ddl}t |j |j|}td |_ d|_tjd y#t$rmddl m } |j||5}|}dddn #1swYnxYwn8#t$r,t |j |j|}YnwxYwYwxYw) zInitialize the Silero VAD analyzer. Args: sample_rate: Audio sample rate (8000 or 16000 Hz). If None, will be set later. params: VAD parameters for detection thresholds and timing. rQzLoading Silero VAD model...zsilero_vad.onnxzpipecat.audio.vad.datarN) resourcesT)rzLoaded Silero VAD)superrrdebugimportlib_resourcesstrfilesjoinpath BaseException importlibrUrr_model_last_reset_time) rrRrS model_name package_path impresourcesmodel_file_pathf __class__s rrzSileroVADAnalyzer.__init__s [@ 23& /  ] 6!,"4"4\"B"K"KJ"WXO&odK ! () ] ; ]!&&|Z@(A&'O(((  ]"%l&8&8&F&O&OPZ&["\ ]  ]sM-B C?C+B7. C7C <CC?2C96C?8C99C?>C?cV|dk7r|dk7rtd|dt| |y)zSet the sample rate for audio processing. Args: sample_rate: Audio sample rate (must be 8000 or 16000 Hz). Raises: ValueError: If sample rate is not 8000 or 16000 Hz. rrz?Silero VAD sample rate needs to be 16000 or 8000 (sample rate: )N)r&rVset_sample_rate)rrRres rrhz!SileroVADAnalyzer.set_sample_rates? % K4$7QR]Q^^_`   ,rreturnc(|jdk(rdSdS)zGet the number of audio frames required for VAD analysis. Returns: Number of frames required (512 for 16kHz, 256 for 8kHz). rr7r8)rR)rs rnum_frames_requiredz%SileroVADAnalyzer.num_frames_requireds &&%/s8S8rc tj|tj}tj|tjjtjdz }|j ||j d}tj}||jz }|tk\r!|j j||_|S#t$r"}tjd|Yd}~yd}~wwxYw)zCalculate voice activity confidence for the given audio buffer. Args: buffer: Audio buffer to analyze. Returns: Voice confidence score between 0.0 and 1.0. r.g@rz'Error analyzing audio with Silero VAD: N)r# frombufferint16astyper-r^rRtimer__MODEL_RESET_STATES_TIMEr Exceptionrerror)rbuffer audio_int16 audio_float32new_confidence curr_time diff_timees rvoice_confidencez"SileroVADAnalyzer.voice_confidences --9KMM+RXXFMMbjjY\ccM![[8H8HI!LN I!D$9$99I44 ((*(1%! !  LLB1#F G sCC D"C??D) rIrJrKrLrrMrrrhrkfloatr{ __classcell__)res@rrPrPsQ8<[_*x}*XiEX*J-3- 9S9%rrP)rLrptypingrnumpyr#logururpipecat.audio.vad.vad_analyzerrrrqrModuleNotFoundErrorrzrsrrrrPrNrrrs A/]]@` `M/FLL;qc"#FLLTU )!- ../s<A82A33A8