Ir para o conteúdo

VoiceHubvad_webrtc

WebRTCVAD

Runs weightless WebRTC VAD with frame-compatible duration controls.

Voice activity detectionVoiceHub-nativewebrtc-vadParameters: WeightlessNot text-language conditionedTraining: inference-onlyLicense: MIT and BSD-3-Clause

Parameter metadata: Weightless algorithm; the registered default has no model parameters.

Use this model
Resources
Upstream GitHub VoiceHub source

Usage

Complete the VoiceHub installation once, then run this repository-authored example. Model pages intentionally contain no package-install command.

This example is maintained against VoiceHub's public API; it is not copied from an upstream demo or package README.

Model-specific path: Runs weightless WebRTC VAD with frame-compatible duration controls.

Inputs and controls: Input is resampled and framed by VoiceHub; algorithm aggressiveness belongs to the model configuration.

from pathlib import Path

from voicehub import AutoModelForVoiceActivityDetection

AUDIO_FILE = Path("speech.wav")
if not AUDIO_FILE.is_file():
    raise FileNotFoundError(AUDIO_FILE)

model = AutoModelForVoiceActivityDetection.from_pretrained(
    'webrtc-vad',
    model_type='vad_webrtc',
    device="cpu",
    lazy_load=True,
)
output = model.detect(
    AUDIO_FILE,
    min_speech_duration_ms=120,
    min_silence_duration_ms=240,
    speech_pad_ms=30,
)
for segment in output.segments:
    print(segment.start, segment.end, segment.score)

Use authorized recordings. Version the implementation and configuration in production.

Overview

vad_webrtc is a VoiceHub voice activity detection integration. This page is generated from its registry contract.

Property Value
Task Voice activity detection
Architecture webrtc-vad
Runtime VoiceHub-native
Languages Not text-language conditioned
Capabilities voice-activity-detection, fixed-point, voicehub-native, native-runtime, streaming
Reusable components —
Normalized output VADOutput

Language support

This weightless runtime does not select a spoken language and is not text-language conditioned; validate its implementation, configuration, and recording conditions for the target speech.

Paper and GitHub

Configuration

Load configuration without constructing the model:

from voicehub import AutoConfig

config = AutoConfig.for_model('vad_webrtc')
print(config.model_type)
Property Value
Canonical model type vad_webrtc
Configuration class WebRTCVADConfig
Architecture class WebRTCVADForVoiceActivityDetection

Processing

Create the registered processor without allocating model weights:

from voicehub import AutoProcessor

processor = AutoProcessor.from_pretrained(
    'webrtc-vad',
    model_type='vad_webrtc',
)
print(type(processor).__name__)

Inference

The Usage example returns VADOutput through AutoModelForVoiceActivityDetection.

Input and output contract

Property Value
Label boundary No verified training dataset contract
Required training inputs —

Use authorized audio and preserve annotation provenance. See the ASR and VAD data workflow.

Training and optimization

Use available_optimization_passes() to discover reversible public passes. Unsupported runtime or hardware fails closed before mutation.

Training contract

Property Value
Support inference-only
Family upstream-native
Recipe single-phase
Default phase default
Runtime identifier webrtc-vad
Native training graph no
Phase Kind Components Required inputs Loss keys
default objective — — loss, total_loss

This integration is inference-only. Choose a verified model from the training matrix.

Checkpoints, provenance, license, and limitations

Property Value
Runtime identifier webrtc-vad
Hugging Face ID Not published / not applicable
Not applicable: WebRTC VAD is a weightless signal-processing algorithm.
Checkpoint status Not applicable; this is a weightless algorithm with no checkpoint
Optional dependency extra Core package
Hardware and runtime Usage selects cpu; verify implementation-specific requirements
Real-checkpoint evidence Not applicable; version the implementation, configuration, and source provenance
Implementation voicehub.models.vad_webrtc.modeling_vad_webrtc.WebRTCVADForVoiceActivityDetection
Configuration voicehub.models.vad_webrtc.configuration_vad_webrtc.WebRTCVADConfig
Source provenance voicehub/architectures/webrtc_vad/SOURCE.json
License MIT and BSD-3-Clause

This weightless runtime has no checkpoint license. Its audited source record declares MIT and BSD-3-Clause; verify those implementation terms.

Confirm the implementation revision, source provenance, access terms, and license.

Limitations

  • Weightless algorithm; the registered default has no model parameters.
  • Validate memory, precision, and optional dependencies on the target system.
  • Public optimizations fail closed when the runtime or hardware cannot satisfy their validation contract; an unavailable pass is not reported as applied.
  • Contract tests do not replace implementation and recording-condition validation.

Public API

Use the stable configuration, processor, and task-model facades below.

Configuration

WebRTCVADConfig

View source

WebRTCVADConfig(**config_kwargs)

Parameters

  • **config_kwargs — Configuration fields validated by WebRTCVADConfig.

Model

WebRTCVADForVoiceActivityDetection

View source

AutoModelForVoiceActivityDetection.from_pretrained(
    pretrained_model_name_or_path,
    *,
    model_type='vad_webrtc',
    config=None,
    **model_kwargs,
)

Parameters

  • pretrained_model_name_or_path — Runtime identifier for this weightless implementation.
  • model_type — Canonical model type; use 'vad_webrtc'.
  • config — Optional preloaded WebRTCVADConfig instance.
  • **model_kwargs — Model-specific loading arguments.
from voicehub import get_model_spec

spec = get_model_spec('vad_webrtc')
print(spec.display_name, spec.task.value)
Purpose Public object
Discover get_model_spec('vad_webrtc')
Load and run AutoModelForVoiceActivityDetection
Configure WebRTCVADConfig
Process AutoProcessor
Model implementation WebRTCVADForVoiceActivityDetection
Normalized output VADOutput
Training contract get_training_spec('vad_webrtc')
Optimization lifecycle available_optimization_passes, apply_optimization_plan, optimization_manifest, restore_optimization_plan

See all model guides, inference, and the training matrix.