huri-module-vad 0.1.0.dev0__py3-none-any.whl

huri_module_vad-0.1.0.dev0.dist-info/METADATA

@@ -0,0 +1,39 @@
+Metadata-Version: 2.4
+Name: huri-module-vad
+Version: 0.1.0.dev0
+Summary: HuRI module: Voice activity detection (WebRTC VAD): microphone frames in, speech segments out
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+Requires-Dist: huri-core[server]<0.5,>=0.4.0.dev0
+Requires-Dist: huri-events<0.2,>=0.1.0.dev0
+Requires-Dist: numpy<2,>=1.26
+Requires-Dist: webrtcvad==2.0.10
+Requires-Dist: setuptools<82
+
+# huri-module-vad
+
+Voice activity detection (WebRTC VAD): microphone frames in, speech segments out.
+
+| Module | Topics | What it does |
+|---|---|---|
+| `vad` | `huri.audio.in` → `huri.voice` | Detects speech in microphone frames (WebRTC VAD) |
+
+```sh
+pip install -e packages/huri-module-vad     # from a huri-modules checkout, next to huri-core and huri-events
+HURI_MODULES=vad ...   # enable it on the server
+```
+
+A distribution of
+[huri-modules](https://github.com/Lucy-Robotics/huri-modules), versioned in its
+`pyproject.toml` (see RELEASING.md there). It depends on `huri-core` and
+`huri-events`, which are not on PyPI: without a checkout, install them from git
+in the same pip command, so pip never looks them up on an index:
+
+```sh
+ref=0.1.0   # a huri-modules release tag
+core=0.4.0  # HURI_CORE_REF in that release's versions.env
+pip install \
+  "huri-core[server] @ git+https://github.com/Lucy-Robotics/huri-core.git@$core" \
+  "huri-events @ git+https://github.com/Lucy-Robotics/huri-modules.git@$ref#subdirectory=packages/huri-events" \
+  "huri-module-vad @ git+https://github.com/Lucy-Robotics/huri-modules.git@$ref#subdirectory=packages/huri-module-vad"
+```

huri_module_vad-0.1.0.dev0.dist-info/RECORD

@@ -0,0 +1,8 @@
+huri_vad/__init__.py,sha256=47DEQpj8HBSa-_TImW-5JCeuQeRkm5NMpJWZG3hSuFU,0
+huri_vad/huri_module.toml,sha256=YtYluQ3RDV7JaLZeU2oGSUria7OpTY86v2qSfPbaLyo,337
+huri_vad/module.py,sha256=XMXHcSU66aqG0MOKkwlhjPAOnITsqkm9aib5w6r_W1w,2352
+huri_module_vad-0.1.0.dev0.dist-info/METADATA,sha256=-CI5OfxKN9crpIUT4T5y6cTDD5yzgII43P2Uh2rYHOU,1592
+huri_module_vad-0.1.0.dev0.dist-info/WHEEL,sha256=YLJXdYXQ2FQ0Uqn2J-6iEIC-3iOey8lH3xCtvFLkd8Q,91
+huri_module_vad-0.1.0.dev0.dist-info/entry_points.txt,sha256=rvYkPeqJUyKUMRG7MYJptdWp7epBANLp6xgIrZC6kQo,41
+huri_module_vad-0.1.0.dev0.dist-info/top_level.txt,sha256=Viv-8JWNDXL5jm86N5Q6e_boQf2Sd3Q6sife2hSxFRM,9
+huri_module_vad-0.1.0.dev0.dist-info/RECORD,,

huri_module_vad-0.1.0.dev0.dist-info/WHEEL

@@ -0,0 +1,5 @@
+Wheel-Version: 1.0
+Generator: setuptools (81.0.0)
+Root-Is-Purelib: true
+Tag: py3-none-any
+

huri_module_vad-0.1.0.dev0.dist-info/entry_points.txt

@@ -0,0 +1,2 @@
+[huri.modules]
+vad = huri_vad.module:VAD

huri_module_vad-0.1.0.dev0.dist-info/top_level.txt

@@ -0,0 +1 @@
+huri_vad

huri_vad/huri_module.toml

@@ -0,0 +1,7 @@
+name = "vad"
+summary = "Detects speech in microphone frames (WebRTC VAD)"
+input = { topic = "huri.audio.in", event = "huri_core.core.events:BytesEvent" }
+output = { topic = "huri.voice", event = "huri_events.speech:Voice" }
+events = ["huri-events>=0.1,<0.2"]
+requires = { gpu = false, actuation = false }
+huri_core = ">=0.4.0.dev0,<0.5"

huri_vad/module.py

@@ -0,0 +1,69 @@
+from typing import Optional
+
+import numpy as np
+import webrtcvad
+from huri_events.speech import Voice
+
+from huri_core.core.events import BytesEvent
+from huri_core.core.module import Module
+
+
+class VAD(Module):
+    """VAD Module
+
+    Detect voice and silence using WebRTC VAD.
+
+    input: huri.audio.in,
+    output: huri.voice
+
+    :vad_agressiveness: from 0 (low) to 3 (high, can distord audio).
+    :silence_duration: how many seconds will a no voice be considered a silence.
+    :sample_rate: size of received chunk of audio. Usually 8000, 16000 or 48000.
+    :block_duration: size of received chunk of audio (in s).
+        Can only be 0.010, 0.020 and 0.030.
+    """
+
+    # Inbound microphone frames travel on their own topic so the TTS-output
+    # "audio" topic (consumed by MOV and the client Sender) never collides
+    # with mic input — otherwise raw mic bytes get echoed back to the client.
+    input_type = "huri.audio.in"
+    output_type = "huri.voice"
+    input_event = BytesEvent
+    output_event = Voice
+
+    def __init__(
+        self,
+        vad_agressiveness: int = 3,
+        silence_duration: float = 3,  # s
+        sample_rate: int = 16000,
+        block_duration: float = 0.020,  # s
+    ):
+        super().__init__()
+
+        if block_duration not in [0.010, 0.020, 0.030]:
+            raise RuntimeError("block duration must be 0.010, 0.020 or 0.030 s")
+
+        self.sample_rate: int = sample_rate
+        self.block_size: int = int(block_duration * sample_rate)  # ms
+
+        self.silence_frames_size: int = int(silence_duration * sample_rate)
+        self.silence_frames_count: int = -1
+
+        self.vad = webrtcvad.Vad(vad_agressiveness)
+
+    async def process(self, data: BytesEvent) -> Optional[Voice]:
+        if self.vad.is_speech(data.data, self.sample_rate) is True:
+            self.silence_frames_count = 0
+
+            audio_array = np.frombuffer(data.data, dtype=np.int16)
+            audio_array_float = audio_array.astype(np.float32) / 32768.0
+
+            return Voice(audio_array_float)
+        else:
+            if self.silence_frames_count != -1:
+                self.silence_frames_count += self.block_size
+            if self.silence_frames_count > self.silence_frames_size:
+                self.silence_frames_count = -1  # sent only once
+
+                return Voice(None)
+            return None