Bugfix: fix process leak in PythonWrapper::InvokeCommandWithArgs

It now waits up to 10 seconds for a graceful exit and falls back on
the equivalent of a SIGKILL. The caller is assumed to have signaled to the
process through `in_cb` that an exit is desired.

Also:

* Fix graceful exit path of transcribe_v2.py.
* Add toggle to enable/disable preview text. It is enabled by default.
* Constrain transcription temperature to 0.0. This keeps latency more
  predictable at the cost of some accuracy.

7 files changed, 72 insertions, 19 deletions

diff --git a/GUI/GUI/GUI/Config.cpp b/GUI/GUI/GUI/Config.cpp
index 456f5d4..db4b184 100644
--- a/GUI/GUI/GUI/Config.cpp
+++ b/GUI/GUI/GUI/Config.cpp
@@ -82,6 +82,7 @@ AppConfig::AppConfig(wxTextCtrl* out)
 	enable_profanity_filter(false),

 	enable_debug_mode(false),

 	reset_on_toggle(true),

+	enable_previews(true),

 	gpu_idx(0),

 	keybind("ctrl+x"),

 

@@ -122,6 +123,7 @@ bool AppConfig::Serialize(const std::filesystem::path& path) {
 	cm.Set("enable_profanity_filter", enable_profanity_filter);

 	cm.Set("enable_debug_mode", enable_debug_mode);

 	cm.Set("reset_on_toggle", reset_on_toggle);

+	cm.Set("enable_previews", enable_previews);

 	cm.Set("gpu_idx", gpu_idx);

 	cm.Set("keybind", keybind);

 

@@ -175,6 +177,7 @@ bool AppConfig::Deserialize(const std::filesystem::path& path) {
 	cm.Get("enable_profanity_filter", c.enable_profanity_filter);

 	cm.Get("enable_debug_mode", c.enable_debug_mode);

 	cm.Get("reset_on_toggle", c.reset_on_toggle);

+	cm.Get("enable_previews", c.enable_previews);

 	cm.Get("gpu_idx", c.gpu_idx);

 	cm.Get("keybind", c.keybind);

 

diff --git a/GUI/GUI/GUI/Config.h b/GUI/GUI/GUI/Config.h
index a6f83e2..4137c59 100644
--- a/GUI/GUI/GUI/Config.h
+++ b/GUI/GUI/GUI/Config.h
@@ -68,6 +68,7 @@ public:
 	bool enable_profanity_filter;

 	bool enable_debug_mode;

 	bool reset_on_toggle;

+	bool enable_previews;

 	int gpu_idx;

 	std::string keybind;

 

diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index 9a781e8..2823437 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -70,6 +70,7 @@ namespace {
         ID_PY_APP_ENABLE_PROFANITY_FILTER,

         ID_PY_APP_ENABLE_DEBUG_MODE,

         ID_PY_APP_RESET_ON_TOGGLE,

+        ID_PY_APP_ENABLE_PREVIEWS,

         ID_PY_APP_ROWS,

         ID_PY_APP_COLS,

         ID_PY_APP_GPU_IDX,

@@ -942,6 +943,17 @@ Frame::Frame()
                 );

                 py_app_reset_on_toggle_ = py_app_reset_on_toggle;

 

+                auto* py_app_enable_previews = new wxCheckBox(py_config_panel,

+                    ID_PY_APP_ENABLE_PREVIEWS, "Enable previews");

+                py_app_enable_previews->SetValue(app_c_->enable_previews);

+                py_app_enable_previews->SetToolTip(

+                    "If checked, audio that has not yet stabilized will also "

+                    "be transcribed and shown. Turn this off if you're on a "

+                    "resource-constrained system or if transcription is "

+                    "running slowly."

+                );

+                py_app_enable_previews_ = py_app_enable_previews;

+

                 // Hack: Add newlines before and after the button text to make

                 // the buttons bigger, and easier to click from inside VR.

                 auto* py_app_start_button = new wxButton(py_config_panel,

@@ -957,6 +969,8 @@ Frame::Frame()
                     /*flags=*/wxEXPAND);

                 sizer->Add(py_app_reset_on_toggle, /*proportion=*/0,

                     /*flags=*/wxEXPAND);

+                sizer->Add(py_app_enable_previews, /*proportion=*/0,

+                    /*flags=*/wxEXPAND);

                 sizer->Add(py_app_enable_browser_src, /*proportion=*/0,

                     /*flags=*/wxEXPAND);

                 sizer->Add(py_app_enable_local_beep, /*proportion=*/0,

@@ -1538,6 +1552,9 @@ void Frame::ApplyConfigToInputFields()
     auto* py_app_reset_on_toggle = static_cast<wxCheckBox*>(FindWindowById(ID_PY_APP_RESET_ON_TOGGLE));

     py_app_reset_on_toggle->SetValue(app_c_->reset_on_toggle);

 

+    auto* py_app_enable_previews = static_cast<wxCheckBox*>(FindWindowById(ID_PY_APP_ENABLE_PREVIEWS));

+    py_app_enable_previews->SetValue(app_c_->enable_previews);

+

     // Unity panel

     auto* unity_assets_path = static_cast<wxDirPickerCtrl*>(FindWindowById(ID_UNITY_ASSETS_FILE_PICKER));

     unity_assets_path->SetPath(app_c_->assets_path);

@@ -1574,7 +1591,6 @@ void Frame::OnExit(wxCloseEvent& event)
 {

     OnAppStop();

     OnUnityAutoRefreshStop();

-    // Allow default close processing to continue.

     event.Skip();

 }

 

@@ -2245,6 +2261,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
     const bool enable_profanity_filter = py_app_enable_profanity_filter_->GetValue();

     const bool enable_debug_mode = py_app_enable_debug_mode_->GetValue();

     const bool reset_on_toggle = py_app_reset_on_toggle_->GetValue();

+    const bool enable_previews = py_app_enable_previews_->GetValue();

 

 	ASSIGN_OR_RETURN_VOID(int, rows, stoiInRange(transcribe_out_, py_app_rows_->GetValue().ToStdString(), "rows", 1, 10));

 	ASSIGN_OR_RETURN_VOID(int, cols, stoiInRange(transcribe_out_, py_app_cols_->GetValue().ToStdString(), "cols", 1, 120));

@@ -2277,6 +2294,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
     app_c_->enable_profanity_filter = enable_profanity_filter;

     app_c_->enable_debug_mode = enable_debug_mode;

     app_c_->reset_on_toggle = reset_on_toggle;

+    app_c_->enable_previews = enable_previews;

     app_c_->gpu_idx = gpu_idx;

     app_c_->keybind = keybind;

     app_c_->Serialize(AppConfig::kConfigPath);

@@ -2306,7 +2324,13 @@ void Frame::OnAppStart(wxCommandEvent& event) {
 			}

 		}

 	};

-    auto in_cb = [&](std::string& in) {};

+    auto in_cb = [&](std::string& in) {

+        if (!run_py_app_) {

+            std::ostringstream oss;

+            oss << "exit" << std::endl;

+            in = oss.str();

+        }

+    };

     auto run_cb = [&]() {

         return run_py_app_;

     };

@@ -2338,7 +2362,7 @@ void Frame::OnAppStop() {
     }

     else {

 		py_app_.wait();

-		Log(transcribe_out_, "Stopped transcription engine\n");

+        Log(transcribe_out_, "Stopped transcription engine\n");

     }

     status = obs_app_.wait_for(std::chrono::seconds(0));

     if (status == std::future_status::ready) {

@@ -2346,7 +2370,7 @@ void Frame::OnAppStop() {
     }

     else {

 		obs_app_.wait();

-		Log(transcribe_out_, "Stopped browser source\n");

+        Log(transcribe_out_, "Stopped browser source\n");

     }

     transcript_.Clear();

 }

diff --git a/GUI/GUI/GUI/Frame.h b/GUI/GUI/GUI/Frame.h
index 2a538b4..7baf77f 100644
--- a/GUI/GUI/GUI/Frame.h
+++ b/GUI/GUI/GUI/Frame.h
@@ -72,6 +72,7 @@ private:
     wxCheckBox* py_app_enable_profanity_filter_;

     wxCheckBox* py_app_enable_debug_mode_;

     wxCheckBox* py_app_reset_on_toggle_;

+    wxCheckBox* py_app_enable_previews_;

     wxCheckBox* unity_clear_osc_;

     wxCheckBox* unity_enable_phonemes_;

 

diff --git a/GUI/GUI/GUI/PythonWrapper.cpp b/GUI/GUI/GUI/PythonWrapper.cpp
index 7141037..29b7d75 100644
--- a/GUI/GUI/GUI/PythonWrapper.cpp
+++ b/GUI/GUI/GUI/PythonWrapper.cpp
@@ -320,7 +320,16 @@ bool PythonWrapper::InvokeCommandWithArgs(const std::string& cmd,
 		}
 	}
 	if (!run_cb()) {
-		return false;
+		DWORD timeout_ms = 1000 * 10;
+		DWORD ret = WaitForSingleObject(pi.hProcess, timeout_ms);
+
+		if (ret == WAIT_TIMEOUT) {
+			std::ostringstream stderr_oss;
+			stderr_oss << "Timed out waiting for graceful exit, killing process";
+			out_cb("", stderr_oss.str());
+
+			TerminateProcess(pi.hProcess, 0);
+		}
 	}
 
 	std::ostringstream stdout_oss, stderr_oss;
diff --git a/Scripts/steamvr.py b/Scripts/steamvr.py
index da07134..3e6c6c9 100644
--- a/Scripts/steamvr.py
+++ b/Scripts/steamvr.py
@@ -19,6 +19,7 @@ class InputEvent:
 def pollButtonPress(
         hand: str = "right",
         button: str = "b",
+        ctrl = None  # ThreadControl object
         ) -> int:
     hands = {}
     hands["left"] = vr.TrackedControllerRole_LeftHand
@@ -31,7 +32,7 @@ def pollButtonPress(
 
     system = None
     first = True
-    while not system:
+    while ctrl.run_app and not system:
         try:
             system = vr.init(vr.VRApplication_Background)
         except Exception as e:
@@ -42,7 +43,7 @@ def pollButtonPress(
     last_packet = 0
     event_high = False
 
-    while True:
+    while ctrl.run_app:
         time.sleep(0.01)
 
         lh_idx = system.getTrackedDeviceIndexForControllerRole(hands[hand])
diff --git a/Scripts/transcribe_v2.py b/Scripts/transcribe_v2.py
index 9b3a404..9e642e9 100644
--- a/Scripts/transcribe_v2.py
+++ b/Scripts/transcribe_v2.py
@@ -428,6 +428,7 @@ class Whisper:
                 audio,
                 language = langcodes.find(self.cfg["language"]).language,
                 vad_filter = True,
+                temperature=0.0,
                 without_timestamps = False)
         res = []
         for s in segments:
@@ -463,9 +464,11 @@ def saveAudio(audio: bytes, path: str):
 
 class VadCommitter:
     def __init__(self,
+            cfg: typing.Dict,
             collector: AudioCollector,
             whisper: Whisper,
             segmenter: AudioSegmenter):
+        self.cfg = cfg
         self.collector = collector
         self.whisper = whisper
         self.segmenter = segmenter
@@ -486,7 +489,7 @@ class VadCommitter:
             for s in segments:
                 print(f"commit segment: {s}")
             delta = ''.join(s.transcript for s in segments)
-            #print(f"delta get: {delta}")
+            print(f"delta get: {delta}")
             audio = self.collector.getAudio()
 
             #ts = datetime.fromtimestamp(self.collector.now() - latency_s)
@@ -494,12 +497,13 @@ class VadCommitter:
             #saveAudio(commit_audio, filename)
 
         preview = ""
-        if has_audio:
-            segments = self.whisper.transcribe(audio)
-            preview = "".join(s.transcript for s in segments)
-        else:
-            #print("VAD detects no audio, skip transcription")
-            self.collector.keepLast(1.0)
+        if self.cfg["enable_previews"]:
+            if has_audio:
+                segments = self.whisper.transcribe(audio)
+                preview = "".join(s.transcript for s in segments)
+            else:
+                #print("VAD detects no audio, skip transcription")
+                self.collector.keepLast(1.0)
 
         return TranscriptCommit(
                 delta,
@@ -556,7 +560,7 @@ def evaluate(cfg,
     collector = CompressingAudioCollector(collector)
     whisper = Whisper(collector, cfg)
     segmenter = AudioSegmenter(min_silence_ms=250)
-    committer = VadCommitter(collector, whisper, segmenter)
+    committer = VadCommitter(cfg, collector, whisper, segmenter)
     transcript = ""
     commits = []
     last_commit_ts = None
@@ -685,10 +689,14 @@ def vrInputThread(ctrl: ThreadControl):
     last_rising = time.time()
     last_medium_press_end = 0
 
-    button_generator = steamvr.pollButtonPress(hand=hand_id, button=button_id)
+    button_generator = steamvr.pollButtonPress(hand=hand_id, button=button_id,
+            ctrl=ctrl)
     while ctrl.run_app:
         time.sleep(0.01)
-        event = next(button_generator)
+        try:
+            event = next(button_generator)
+        except StopIteration:
+            break
 
         if event.opcode == steamvr.EVENT_RISING_EDGE:
             last_rising = time.time()
@@ -790,7 +798,7 @@ def kbInputThread(ctrl: ThreadControl):
     PAUSE_STATE = 1
     state = PAUSE_STATE
 
-    while ctrl.run_app == True:
+    while ctrl.run_app:
         time.sleep(0.01)
 
         cur_press_time = machine.getNextPressTime()
@@ -879,7 +887,7 @@ def run(cfg):
     collector = CompressingAudioCollector(collector)
     whisper = Whisper(collector, cfg)
     segmenter = AudioSegmenter(min_silence_ms=250)
-    committer = VadCommitter(collector, whisper, segmenter)
+    committer = VadCommitter(cfg, collector, whisper, segmenter)
     pager = OscPager(cfg)
 
     ctrl = ThreadControl(cfg)
@@ -909,13 +917,19 @@ def run(cfg):
 
     for line in sys.stdin:
         if "exit" in line or "quit" in line:
+            print("Exit requested", file=sys.stderr)
             break
 
     ctrl.run_app = False
+    print("Join transcription thread")
     transcribe_audio_thd.join()
+    print("Join vr input thread")
     vr_input_thd.join()
+    print("Join kb input thread")
     kb_input_thd.join()
+    print("Join osc thread")
     osc_thd.join()
+    print("Done")
 
 if __name__ == "__main__":
     parser = argparse.ArgumentParser()