python-voiceio 0.9.2__py3-none-any.whl → 0.9.4__py3-none-any.whl

{python_voiceio-0.9.2.dist-info → python_voiceio-0.9.4.dist-info}/METADATA

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: python-voiceio
-Version: 0.9.2
+Version: 0.9.4
 Summary: Voice dictation for Linux. Speak → text, locally, instantly.
 Author: Hugo Montenegro
 License-Expression: MIT

{python_voiceio-0.9.2.dist-info → python_voiceio-0.9.4.dist-info}/RECORD

@@ -1,7 +1,7 @@
-python_voiceio-0.9.2.dist-info/licenses/LICENSE,sha256=Gz61o8jFTAvZUZyB3nWDB3DQQVuipjfPkVu9W8hBHM0,1072
-voiceio/__init__.py,sha256=gqT-BGoeEItda9fICQDvLbxEjWRIBhFJxPxxKvmHLUo,22
+python_voiceio-0.9.4.dist-info/licenses/LICENSE,sha256=Gz61o8jFTAvZUZyB3nWDB3DQQVuipjfPkVu9W8hBHM0,1072
+voiceio/__init__.py,sha256=e56AvHfJCtG2ZwwINqsxINVbehWdKxMYgIDbjd7P-II,22
 voiceio/__main__.py,sha256=xT5QCGGreYMisHO7Lh_Y-xAQ2TOzG2D7npKknGNcWSY,53
-voiceio/app.py,sha256=ScrUVDQQ_YbAFKKSMOhaRPX_WTcPuenhnRdSzWMxThE,73785
+voiceio/app.py,sha256=8xS_MySy6cvpxXDcd0-6YwQK6KW_iT8Vdf6FqzYUXpI,74743
 voiceio/audit.py,sha256=p4F2BBX4wEC9eS8MSplzfXL52XC2dyIKbgrRvR-WbTU,15802
 voiceio/autocorrect.py,sha256=jGsglEYwAHWk2uWedeEZzyAMW2LWO_8nQGZgrWYy4Dc,26552
 voiceio/autocorrect_state.py,sha256=BNzdHP5slcOV_-OU4MlaKE7xgVJzrPncsYFK30JWTs8,6858
@@ -9,7 +9,7 @@ voiceio/backends.py,sha256=-rs1YhPblub4NjFb6qCD7Ih9FqBRXtPuvjVv9bQoQ1w,368
 voiceio/cli.py,sha256=QE8N7N3d1O7r9-u_jbYMAH9gw0ILX7dSxDZZYj83lIs,78969
 voiceio/clipboard_read.py,sha256=lz75CkRGh7pyvO6xvoUmxJDl-Gwie1xlXQHK70PO84I,4261
 voiceio/commands.py,sha256=Vhtn8s5G5OcCbRN44YZ-_14Fa85ISXl_3Ix-KqSHs2A,4694
-voiceio/config.py,sha256=DA1PGIanhrPjWkVT6jj0NSFslW9vuegzV6eMJNuKBzM,15273
+voiceio/config.py,sha256=IxXfs780V4eMhbfoyHZAUER7cmjg950Lp1wsvm9aEIU,15563
 voiceio/consent.py,sha256=jtdp1TkTbicUSTCDmV4qo4SdbS4kQpYzy1Kjc_7pJVw,1883
 voiceio/corrections.py,sha256=7moKhkVvo5CZU1mZobV2-MajkcHiwliKBY1dSBFh0w0,7445
 voiceio/demo.py,sha256=QRNJuNObXCsjwGD-Cz3dh81cfb21xiuSeVyyU720U9Q,8673
@@ -26,7 +26,7 @@ voiceio/platform.py,sha256=f_8MmC5q1sCEuEPyKjRqlVvV9FExuGmwn_aJXhkpiEk,11809
 voiceio/postcorrect.py,sha256=iyO7MPIR4a4XIASEGOs9Y3iUXVTfwtVkhVOWj2vY1lk,11201
 voiceio/postprocess.py,sha256=9grkq3I6XgiQV6ZQbIw2_fOUe57tzC7GBPvLXqhhdTs,2979
 voiceio/prompt.py,sha256=9iza8KoQxvO_U9NPMJ8Ep7j-zkIs9nW8N5sacCRpsEk,2904
-voiceio/recorder.py,sha256=p1r0mH_plBD3BHqBbIGV_m2N4Mj5RAF7LfMu8u6mk5I,12813
+voiceio/recorder.py,sha256=csqm4Q_GuQyDxhJAVDCL6TQd_nna77sfWNiwqF6TsQk,15003
 voiceio/retention.py,sha256=i_VWPCyykT4kQ1uKilTplpiIrACTYljn09BU1vL71A0,6019
 voiceio/service.py,sha256=1YhnDPHnu0Aw66uwsp05TWfs18bf3io_h6r8eNla7nA,12525
 voiceio/snapshots.py,sha256=bm-lzIeYucDUDlPBZiFiWd-jyB2I6Ft1jRANZCO6go4,2351
@@ -74,8 +74,8 @@ voiceio/typers/pynput_type.py,sha256=DTtkT59M-EKOAlTNkvcZrRvf_Fq52OpI8S5t_8S-QAY
 voiceio/typers/wtype.py,sha256=d1wG-HZdYDQxXIysV_Xvk9HixDpaJgl3VEoKbP5vhIs,1786
 voiceio/typers/xdotool.py,sha256=dh1zhPqUT8ihJahJ7ZKm6PtfYY087UAzADx664DvQOM,1356
 voiceio/typers/ydotool.py,sha256=dt0W9ot__W8LQG2so8vZLGIw9leK2Q6g3i-Voo0LXE4,5327
-python_voiceio-0.9.2.dist-info/METADATA,sha256=RZm3kWz94AUBzUicA4CWYbm_oLyaFrNVfvU0acqEHEY,15877
-python_voiceio-0.9.2.dist-info/WHEEL,sha256=K260EYznzXsJYBQGqmI8VTxEdiZYNvDZwW9cBh9-_MA,91
-python_voiceio-0.9.2.dist-info/entry_points.txt,sha256=U64fA65zxzyLoC8bgbn2ztQVWHLsc0o0H0qCk1J9DMc,218
-python_voiceio-0.9.2.dist-info/top_level.txt,sha256=piwtn309lD6uexQyXdZ-efAVBJF9y6Wfr48Z-8zkNhg,8
-python_voiceio-0.9.2.dist-info/RECORD,,
+python_voiceio-0.9.4.dist-info/METADATA,sha256=jwrZH_VW6__XC_VKHFMHr052Iadetbx3V5Ma26UYZNg,15877
+python_voiceio-0.9.4.dist-info/WHEEL,sha256=K260EYznzXsJYBQGqmI8VTxEdiZYNvDZwW9cBh9-_MA,91
+python_voiceio-0.9.4.dist-info/entry_points.txt,sha256=U64fA65zxzyLoC8bgbn2ztQVWHLsc0o0H0qCk1J9DMc,218
+python_voiceio-0.9.4.dist-info/top_level.txt,sha256=piwtn309lD6uexQyXdZ-efAVBJF9y6Wfr48Z-8zkNhg,8
+python_voiceio-0.9.4.dist-info/RECORD,,

voiceio/__init__.py

@@ -1 +1 @@
-__version__ = "0.9.2"
+__version__ = "0.9.4"

voiceio/app.py

@@ -283,8 +283,28 @@ class VoiceIO:
             self._last_hotkey = now
             self._toggle()
 
-    def _on_auto_stop(self) -> None:
-        """Called from audio thread when sustained silence triggers auto-stop."""
+    def _on_auto_stop(self, reason: str = "silence") -> None:
+        """Called from the audio thread when the recorder auto-stops.
+
+        reason="silence": spoke, then went quiet — a normal end of dictation.
+        reason="no_speech": nothing was heard the whole time — a muted mic or a
+        hotkey pressed by accident. Surface it so it isn't a silent failure
+        (today's Discord-mute confusion: recording ran, nothing was captured,
+        and there was no sign why).
+        """
+        if reason == "no_speech":
+            from voiceio import feedback
+            if not self.recorder.has_signal():
+                feedback.notify(
+                    "VoiceIO: microphone is muted",
+                    "Stopped — no audio is reaching voiceio. Check your mic's "
+                    "mute button and that the right input device is selected.",
+                )
+            else:
+                feedback.notify(
+                    "VoiceIO: stopped — no speech detected",
+                    "Recording auto-stopped after silence.",
+                )
         threading.Thread(target=self._request_stop, daemon=True).start()
 
     def _request_stop(self) -> None:

voiceio/config.py

@@ -71,6 +71,11 @@ class AudioConfig:
     silence_threshold: float = 0.01
     silence_duration: float = 0.6
     auto_stop_silence_secs: float = 5.0
+    # Safety net for a muted mic or a forgotten hotkey: if NO speech is heard
+    # for this long from the start, stop and notify. The 5s auto-stop above
+    # only fires once speech has been heard, so it can't catch a dead mic.
+    # 0 disables.
+    auto_stop_no_speech_secs: float = 20.0
     vad_backend: str = "silero"  # "silero" | "rms"
     vad_threshold: float = 0.5  # Silero speech probability threshold
 

voiceio/recorder.py

@@ -15,6 +15,11 @@ if TYPE_CHECKING:
 
 log = logging.getLogger(__name__)
 
+# Peak amplitude below this is treated as no signal at all (a muted mic emits
+# exact zeros; a working mic's "silence" still noises above ~1e-4). Matches the
+# floor in has_signal(). Used only to detect a dead mic, never speech vs pause.
+_SIGNAL_FLOOR = 1e-4
+
 
 class RingBuffer:
     """Fixed-size ring buffer for float32 audio samples."""
@@ -108,9 +113,16 @@ class AudioRecorder:
 
         # Auto-stop on sustained silence
         self._auto_stop_secs = cfg.auto_stop_silence_secs
+        self._no_speech_secs = cfg.auto_stop_no_speech_secs
         self._sustained_silence = 0.0
         self._heard_speech = False
-        self._on_auto_stop: Callable[[], None] | None = None
+        # Raw-signal (amplitude) tracking for the muted-mic auto-stop, kept
+        # separate from the VAD-based _heard_speech / _sustained_silence above.
+        self._heard_signal = False
+        self._silent_signal_secs = 0.0
+        # Called with a reason: "silence" (spoke then paused) or "no_speech"
+        # (nothing ever heard — a muted mic or a forgotten hotkey).
+        self._on_auto_stop: Callable[[str], None] | None = None
 
         # Heartbeat: updated by _callback, checked by health watchdog
         self._last_callback_time: float = 0.0
@@ -218,6 +230,8 @@ class AudioRecorder:
             self._silent_chunks = 0.0
             self._sustained_silence = 0.0
             self._heard_speech = False
+            self._heard_signal = False
+            self._silent_signal_secs = 0.0
             self._pause_fired_at = 0
             self._meter_peak = 0.0
             self._meter_clipped = 0
@@ -260,7 +274,7 @@ class AudioRecorder:
         """Set/clear the speech pause callback (used by streaming session)."""
         self._on_speech_pause = callback
 
-    def set_on_auto_stop(self, callback: Callable[[], None] | None) -> None:
+    def set_on_auto_stop(self, callback: Callable[[str], None] | None) -> None:
         """Set/clear the auto-stop callback (fires after sustained silence)."""
         self._on_auto_stop = callback
 
@@ -298,8 +312,19 @@ class AudioRecorder:
 
         # Level metering
         abs_chunk = np.abs(chunk.ravel())
-        self._meter_peak = max(self._meter_peak, float(abs_chunk.max(initial=0.0)))
+        chunk_peak = float(abs_chunk.max(initial=0.0))
+        self._meter_peak = max(self._meter_peak, chunk_peak)
         self._meter_samples += len(abs_chunk)
+
+        # Raw-signal presence, independent of the VAD. A muted mic delivers
+        # digital zeros; real audio (even speech the VAD fails to classify)
+        # sits well above the floor. This — NOT the VAD — gates the muted-mic
+        # auto-stop, so a recording is never cut while real audio is coming in.
+        if chunk_peak > _SIGNAL_FLOOR:
+            self._heard_signal = True
+            self._silent_signal_secs = 0.0
+        else:
+            self._silent_signal_secs += frames / self.sample_rate
         pinned = abs_chunk >= 0.99
         if np.count_nonzero(pinned) >= 4:
             runs = np.convolve(pinned.astype(np.int8), np.ones(4, dtype=np.int8), "valid")
@@ -340,4 +365,20 @@ class AudioRecorder:
             self._on_auto_stop = None
             self._sustained_silence = 0.0
             log.info("Auto-stopping after %.0fs of silence", self._auto_stop_secs)
-            cb()
+            cb("silence")
+
+        # Safety net for a dead mic: NO raw audio signal at all for a long
+        # time — the mic is muted / unplugged / delivering zeros, never merely
+        # that the VAD didn't flag speech (that would cut real dictation off
+        # mid-sentence). _heard_signal latches on the first non-zero chunk, so
+        # this only fires for a mic that was silent from the very start.
+        elif (self._on_auto_stop is not None
+                and self._no_speech_secs > 0
+                and not self._heard_signal
+                and self._silent_signal_secs >= self._no_speech_secs):
+            cb = self._on_auto_stop
+            self._on_auto_stop = None
+            self._silent_signal_secs = 0.0
+            log.info("Auto-stopping: no mic signal in %.0fs (muted?)",
+                     self._no_speech_secs)
+            cb("no_speech")