diff --git a/docs/api/processors/pyhealth.processors.SenselabEGeMAPSProcessor.rst b/docs/api/processors/pyhealth.processors.SenselabEGeMAPSProcessor.rst
new file mode 100644
index 000000000..0ae85746d
--- /dev/null
+++ b/docs/api/processors/pyhealth.processors.SenselabEGeMAPSProcessor.rst
@@ -0,0 +1,7 @@
+pyhealth.processors.SenselabEGeMAPSProcessor
+=============================================
+
+.. autoclass:: pyhealth.processors.SenselabEGeMAPSProcessor
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/api/processors/pyhealth.processors.SenselabEmbeddingProcessor.rst b/docs/api/processors/pyhealth.processors.SenselabEmbeddingProcessor.rst
new file mode 100644
index 000000000..caaead3ae
--- /dev/null
+++ b/docs/api/processors/pyhealth.processors.SenselabEmbeddingProcessor.rst
@@ -0,0 +1,7 @@
+pyhealth.processors.SenselabEmbeddingProcessor
+===============================================
+
+.. autoclass:: pyhealth.processors.SenselabEmbeddingProcessor
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/docs/api/processors/pyhealth.processors.SenselabVADProcessor.rst b/docs/api/processors/pyhealth.processors.SenselabVADProcessor.rst
new file mode 100644
index 000000000..0c6926adc
--- /dev/null
+++ b/docs/api/processors/pyhealth.processors.SenselabVADProcessor.rst
@@ -0,0 +1,7 @@
+pyhealth.processors.SenselabVADProcessor
+=========================================
+
+.. autoclass:: pyhealth.processors.SenselabVADProcessor
+   :members:
+   :undoc-members:
+   :show-inheritance:
diff --git a/examples/senselab_processor_tutorial.py b/examples/senselab_processor_tutorial.py
new file mode 100644
index 000000000..5cebcc8df
--- /dev/null
+++ b/examples/senselab_processor_tutorial.py
@@ -0,0 +1,70 @@
+"""
+Senselab Processor Tutorial
+============================
+Demonstrates PyHealth's senselab processor wrappers for voice
+biomarker extraction: VAD, eGeMAPS features, and speaker embeddings.
+
+Use cases include depression detection, Parkinson's monitoring,
+and cognitive assessment through voice.
+
+Requires: pip install senselab
+"""
+import os
+import tempfile
+
+import torch
+import torchaudio
+
+
+def make_test_wav(path: str, sr: int = 16000, duration: float = 3.0) -> None:
+    """Write a synthetic sine-wave audio file for demonstration."""
+    t = torch.linspace(0, duration, int(sr * duration))
+    waveform = torch.sin(2 * torch.pi * 440 * t).unsqueeze(0)
+    torchaudio.save(path, waveform, sr)
+
+
+with tempfile.TemporaryDirectory() as tmpdir:
+    audio_path = os.path.join(tmpdir, "sample.wav")
+    make_test_wav(audio_path)
+
+    # ------------------------------------------------------------------
+    # 1. SenselabVADProcessor
+    #    Strips silence and noise, returns speech-only waveform tensor.
+    # ------------------------------------------------------------------
+    from pyhealth.processors import SenselabVADProcessor
+
+    vad = SenselabVADProcessor(sample_rate=16000)
+    speech = vad.process(audio_path)
+    print(f"VAD output shape:     {speech.shape}")
+    # Expected: torch.Size([1, N]) where N <= original samples
+
+    # ------------------------------------------------------------------
+    # 2. SenselabEGeMAPSProcessor
+    #    Extracts 88 hand-crafted eGeMAPS acoustic features widely used
+    #    in clinical voice research.
+    # ------------------------------------------------------------------
+    from pyhealth.processors import SenselabEGeMAPSProcessor
+
+    egemaps = SenselabEGeMAPSProcessor(sample_rate=16000, apply_vad=True)
+    features = egemaps.process(audio_path)
+    print(f"eGeMAPS output shape: {features.shape}")
+    # Expected: torch.Size([88])
+
+    # ------------------------------------------------------------------
+    # 3. SenselabEmbeddingProcessor
+    #    Extracts deep neural speaker embeddings using WavLM or
+    #    ECAPA-TDNN for voice biomarker tasks.
+    # ------------------------------------------------------------------
+    from pyhealth.processors import SenselabEmbeddingProcessor
+
+    # WavLM embedding (768-dim)
+    wavlm = SenselabEmbeddingProcessor(model="wavlm", sample_rate=16000)
+    embedding = wavlm.process(audio_path)
+    print(f"WavLM embedding shape: {embedding.shape}")
+    # Expected: torch.Size([768])
+
+    # ECAPA-TDNN embedding (192-dim)
+    ecapa = SenselabEmbeddingProcessor(model="ecapa", sample_rate=16000)
+    embedding_ecapa = ecapa.process(audio_path)
+    print(f"ECAPA embedding shape: {embedding_ecapa.shape}")
+    # Expected: torch.Size([192])
diff --git a/pyhealth/processors/__init__.py b/pyhealth/processors/__init__.py
index 4568a5ece..cd5e9e2d0 100644
--- a/pyhealth/processors/__init__.py
+++ b/pyhealth/processors/__init__.py
@@ -18,6 +18,11 @@ def get_processor(name: str):
 
 
 # Import all processors so they register themselves
+from .senselab_processor import (
+    SenselabVADProcessor,
+    SenselabEGeMAPSProcessor,
+    SenselabEmbeddingProcessor,
+)
 from .image_processor import ImageProcessor
 from .label_processor import (
     BinaryLabelProcessor,
@@ -82,4 +87,7 @@ def get_processor(name: str):
     "TupleTimeTextProcessor",
     "CehrProcessor",
     "ConceptVocab",
+    "SenselabVADProcessor",
+    "SenselabEGeMAPSProcessor",
+    "SenselabEmbeddingProcessor",
 ]
diff --git a/pyhealth/processors/senselab_processor.py b/pyhealth/processors/senselab_processor.py
new file mode 100644
index 000000000..a37b292ca
--- /dev/null
+++ b/pyhealth/processors/senselab_processor.py
@@ -0,0 +1,290 @@
+from pathlib import Path
+from typing import Any, Literal, Optional, Union
+
+import torch
+
+from . import register_processor
+from .base_processor import FeatureProcessor
+
+
+@register_processor("senselab_vad")
+class SenselabVADProcessor(FeatureProcessor):
+    """Voice Activity Detection (VAD) processor using senselab.
+
+    Removes non-speech segments from an audio file before returning
+    a waveform tensor. Useful as a preprocessing step before feature
+    extraction in voice biomarker pipelines (e.g. depression detection,
+    Parkinson's monitoring).
+
+    Args:
+        sample_rate: Target sample rate in Hz. Defaults to 16000
+            (required by most senselab VAD models).
+        return_tensor: If True, returns a torch.Tensor (channels, samples).
+            If False, returns a list of senselab Audio segments.
+            Defaults to True.
+
+    Raises:
+        ImportError: If senselab is not installed.
+        FileNotFoundError: If the audio file path does not exist.
+
+    Example:
+        >>> processor = SenselabVADProcessor(sample_rate=16000)
+        >>> tensor = processor.process("path/to/audio.wav")
+        >>> tensor.shape
+        torch.Size([1, N])
+    """
+
+    def __init__(
+        self,
+        sample_rate: int = 16000,
+        return_tensor: bool = True,
+    ) -> None:
+        try:
+            import senselab  # noqa: F401
+        except ImportError:
+            raise ImportError(
+                "SenselabVADProcessor requires senselab. "
+                "Install it with: pip install senselab"
+            )
+        self.sample_rate = sample_rate
+        self.return_tensor = return_tensor
+
+    def process(self, value: Union[str, Path]) -> Any:
+        """Run VAD on a single audio file and return speech-only audio.
+
+        Args:
+            value: Path to the audio file.
+
+        Returns:
+            torch.Tensor of shape (1, N) containing concatenated speech
+            segments, or a list of senselab Audio objects if
+            return_tensor=False.
+
+        Raises:
+            FileNotFoundError: If the file does not exist.
+        """
+        import torchaudio
+        from senselab.audio.data_structures import Audio
+        from senselab.audio.tasks.preprocessing.preprocessing import (
+            resample_audios,
+        )
+        from senselab.audio.tasks.voice_activity_detection.api import (
+            detect_voice_activity_in_audios,
+        )
+
+        audio_path = Path(value)
+        if not audio_path.exists():
+            raise FileNotFoundError(f"Audio file not found: {audio_path}")
+
+        waveform, orig_sr = torchaudio.load(audio_path)
+        audio = Audio(waveform=waveform, sampling_rate=orig_sr)
+
+        # Resample to target rate
+        [audio] = resample_audios([audio], resample_rate=self.sample_rate)
+
+        # Run VAD — returns list of Audio objects (speech segments only)
+        speech_segments = detect_voice_activity_in_audios([audio])
+
+        if not self.return_tensor:
+            return speech_segments
+
+        if not speech_segments:
+            return torch.zeros(1, 0)
+
+        waveforms = [seg.waveform for seg in speech_segments]
+        combined = torch.cat(waveforms, dim=-1)
+        return combined
+
+    def is_token(self) -> bool:
+        return False
+
+
+@register_processor("senselab_egemaps")
+class SenselabEGeMAPSProcessor(FeatureProcessor):
+    """eGeMAPS feature extraction processor using senselab.
+
+    Extracts the extended Geneva Minimalistic Acoustic Parameter Set
+    (eGeMAPS) — 88 hand-crafted acoustic features widely used in
+    clinical voice research for conditions such as depression,
+    Parkinson's disease, and cognitive assessment.
+
+    Args:
+        sample_rate: Target sample rate in Hz. Defaults to 16000.
+        apply_vad: Whether to apply Voice Activity Detection before
+            extraction. Defaults to True.
+
+    Raises:
+        ImportError: If senselab is not installed.
+        FileNotFoundError: If the audio file path does not exist.
+
+    Example:
+        >>> processor = SenselabEGeMAPSProcessor()
+        >>> features = processor.process("path/to/audio.wav")
+        >>> features.shape
+        torch.Size([88])
+    """
+
+    def __init__(
+        self,
+        sample_rate: int = 16000,
+        apply_vad: bool = True,
+    ) -> None:
+        try:
+            import senselab  # noqa: F401
+        except ImportError:
+            raise ImportError(
+                "SenselabEGeMAPSProcessor requires senselab. "
+                "Install it with: pip install senselab"
+            )
+        self.sample_rate = sample_rate
+        self.apply_vad = apply_vad
+
+    def process(self, value: Union[str, Path]) -> torch.Tensor:
+        """Extract eGeMAPS features from a single audio file.
+
+        Args:
+            value: Path to the audio file.
+
+        Returns:
+            torch.Tensor of shape (88,) containing eGeMAPS features.
+
+        Raises:
+            FileNotFoundError: If the file does not exist.
+        """
+        import torchaudio
+        from senselab.audio.data_structures import Audio
+        from senselab.audio.tasks.features_extraction.opensmile import (
+            extract_features_from_audios_with_opensmile,
+        )
+        from senselab.audio.tasks.preprocessing.preprocessing import (
+            resample_audios,
+        )
+        from senselab.audio.tasks.voice_activity_detection.api import (
+            detect_voice_activity_in_audios,
+        )
+
+        audio_path = Path(value)
+        if not audio_path.exists():
+            raise FileNotFoundError(f"Audio file not found: {audio_path}")
+
+        waveform, orig_sr = torchaudio.load(audio_path)
+        audio = Audio(waveform=waveform, sampling_rate=orig_sr)
+        [audio] = resample_audios([audio], resample_rate=self.sample_rate)
+
+        if self.apply_vad:
+            segments = detect_voice_activity_in_audios([audio])
+            if segments:
+                combined = torch.cat([s.waveform for s in segments], dim=-1)
+                audio = Audio(waveform=combined, sampling_rate=self.sample_rate)
+
+        features_df = extract_features_from_audios_with_opensmile([audio])
+        feature_tensor = torch.tensor(
+            features_df.values[0], dtype=torch.float32
+        )
+        return feature_tensor
+
+    def is_token(self) -> bool:
+        return False
+
+
+@register_processor("senselab_embedding")
+class SenselabEmbeddingProcessor(FeatureProcessor):
+    """Deep speech embedding processor using senselab (WavLM / ECAPA-TDNN).
+
+    Extracts dense neural embeddings from speech audio using pretrained
+    models. These embeddings capture speaker identity and paralinguistic
+    features useful for voice biomarker tasks.
+
+    Args:
+        model: Embedding model to use. One of ``"wavlm"`` or ``"ecapa"``.
+            Defaults to ``"wavlm"``.
+        sample_rate: Target sample rate in Hz. Defaults to 16000.
+        apply_vad: Whether to apply VAD before embedding. Defaults to True.
+
+    Raises:
+        ImportError: If senselab is not installed.
+        ValueError: If an unsupported model name is provided.
+        FileNotFoundError: If the audio file path does not exist.
+
+    Example:
+        >>> processor = SenselabEmbeddingProcessor(model="wavlm")
+        >>> embedding = processor.process("path/to/audio.wav")
+        >>> embedding.shape
+        torch.Size([768])
+    """
+
+    SUPPORTED_MODELS = ("wavlm", "ecapa")
+
+    def __init__(
+        self,
+        model: Literal["wavlm", "ecapa"] = "wavlm",
+        sample_rate: int = 16000,
+        apply_vad: bool = True,
+    ) -> None:
+        try:
+            import senselab  # noqa: F401
+        except ImportError:
+            raise ImportError(
+                "SenselabEmbeddingProcessor requires senselab. "
+                "Install it with: pip install senselab"
+            )
+        if model not in self.SUPPORTED_MODELS:
+            raise ValueError(
+                f"Unsupported model '{model}'. "
+                f"Choose from: {self.SUPPORTED_MODELS}"
+            )
+        self.model = model
+        self.sample_rate = sample_rate
+        self.apply_vad = apply_vad
+
+    def process(self, value: Union[str, Path]) -> torch.Tensor:
+        """Extract a deep embedding from a single audio file.
+
+        Args:
+            value: Path to the audio file.
+
+        Returns:
+            torch.Tensor of shape (D,) where D is the embedding dimension
+            of the chosen model (768 for WavLM, 192 for ECAPA-TDNN).
+
+        Raises:
+            FileNotFoundError: If the file does not exist.
+        """
+        import torchaudio
+        from senselab.audio.data_structures import Audio
+        from senselab.audio.tasks.preprocessing.preprocessing import (
+            resample_audios,
+        )
+        from senselab.audio.tasks.speaker_embeddings.api import (
+            extract_speaker_embeddings_from_audios,
+        )
+        from senselab.audio.tasks.voice_activity_detection.api import (
+            detect_voice_activity_in_audios,
+        )
+
+        audio_path = Path(value)
+        if not audio_path.exists():
+            raise FileNotFoundError(f"Audio file not found: {audio_path}")
+
+        waveform, orig_sr = torchaudio.load(audio_path)
+        audio = Audio(waveform=waveform, sampling_rate=orig_sr)
+        [audio] = resample_audios([audio], resample_rate=self.sample_rate)
+
+        if self.apply_vad:
+            segments = detect_voice_activity_in_audios([audio])
+            if segments:
+                combined = torch.cat([s.waveform for s in segments], dim=-1)
+                audio = Audio(waveform=combined, sampling_rate=self.sample_rate)
+
+        if self.model == "wavlm":
+            model_spec = {"path": "microsoft/wavlm-base-plus"}
+        else:
+            model_spec = {"path": "speechbrain/spkrec-ecapa-voxceleb"}
+
+        embeddings = extract_speaker_embeddings_from_audios(
+            [audio], model=model_spec
+        )
+        return embeddings[0].squeeze(0)
+
+    def is_token(self) -> bool:
+        return False
diff --git a/tests/processors/test_senselab_processor.py b/tests/processors/test_senselab_processor.py
new file mode 100644
index 000000000..670d0ab65
--- /dev/null
+++ b/tests/processors/test_senselab_processor.py
@@ -0,0 +1,98 @@
+"""Tests for senselab-based processors.
+
+These tests use synthetic audio (generated with torch) to avoid
+requiring real audio files or network downloads.
+"""
+import pytest
+import torch
+
+
+def make_sine_wav(tmp_path, sr=16000, duration=2.0):
+    """Write a sine wave to a temp WAV file and return its path."""
+    import torchaudio
+    t = torch.linspace(0, duration, int(sr * duration))
+    waveform = torch.sin(2 * torch.pi * 440 * t).unsqueeze(0)
+    path = tmp_path / "test.wav"
+    torchaudio.save(str(path), waveform, sr)
+    return path
+
+
+# --- SenselabVADProcessor ---
+
+def test_vad_processor_returns_tensor(tmp_path):
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabVADProcessor
+
+    p = SenselabVADProcessor()
+    wav = make_sine_wav(tmp_path)
+    result = p.process(str(wav))
+    assert isinstance(result, torch.Tensor)
+    assert result.ndim == 2  # (channels, samples)
+
+
+def test_vad_processor_file_not_found():
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabVADProcessor
+
+    p = SenselabVADProcessor()
+    with pytest.raises(FileNotFoundError):
+        p.process("/nonexistent/audio.wav")
+
+
+def test_vad_processor_is_not_token():
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabVADProcessor
+
+    assert SenselabVADProcessor().is_token() is False
+
+
+# --- SenselabEGeMAPSProcessor ---
+
+def test_egemaps_processor_output_shape(tmp_path):
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabEGeMAPSProcessor
+
+    p = SenselabEGeMAPSProcessor()
+    wav = make_sine_wav(tmp_path)
+    result = p.process(str(wav))
+    assert isinstance(result, torch.Tensor)
+    assert result.ndim == 1
+    assert result.shape[0] == 88  # eGeMAPS has 88 features
+
+
+def test_egemaps_processor_is_not_token():
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabEGeMAPSProcessor
+
+    assert SenselabEGeMAPSProcessor().is_not_token() is False
+
+
+# --- SenselabEmbeddingProcessor ---
+
+def test_embedding_processor_wavlm_shape(tmp_path):
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabEmbeddingProcessor
+
+    p = SenselabEmbeddingProcessor(model="wavlm")
+    wav = make_sine_wav(tmp_path)
+    result = p.process(str(wav))
+    assert isinstance(result, torch.Tensor)
+    assert result.ndim == 1
+
+
+def test_embedding_processor_invalid_model():
+    pytest.importorskip("senselab")
+    from pyhealth.processors.senselab_processor import SenselabEmbeddingProcessor
+
+    with pytest.raises(ValueError, match="Unsupported model"):
+        SenselabEmbeddingProcessor(model="gpt5_audio")
+
+
+def test_embedding_processor_import_error(monkeypatch):
+    import sys
+    monkeypatch.setitem(sys.modules, "senselab", None)
+    if "pyhealth.processors.senselab_processor" in sys.modules:
+        del sys.modules["pyhealth.processors.senselab_processor"]
+    with pytest.raises(ImportError, match="pip install senselab"):
+        from pyhealth.processors.senselab_processor import SenselabEmbeddingProcessor
+        SenselabEmbeddingProcessor()