summaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
Diffstat
-rw-r--r--GUI/GUI/GUI/Config.cpp3
-rw-r--r--GUI/GUI/GUI/Config.h1
-rw-r--r--GUI/GUI/GUI/Frame.cpp31
-rw-r--r--GUI/GUI/GUI/Frame.h1
-rw-r--r--GUI/GUI/GUI/PythonWrapper.cpp1
-rw-r--r--Scripts/steamvr.py27
-rw-r--r--Scripts/transcribe.py21
7 files changed, 70 insertions, 15 deletions
diff --git a/GUI/GUI/GUI/Config.cpp b/GUI/GUI/GUI/Config.cpp
index 4e6eb48..3994e62 100644
--- a/GUI/GUI/GUI/Config.cpp
+++ b/GUI/GUI/GUI/Config.cpp
@@ -73,6 +73,7 @@ TranscriptionAppConfig::TranscriptionAppConfig()
model("base.en"),
chars_per_sync("20"),
bytes_per_char("1"),
+ button("left joystick"),
rows("4"),
cols("48"),
window_duration("15"),
@@ -90,6 +91,7 @@ bool TranscriptionAppConfig::Serialize(const std::filesystem::path& path) {
root["model"] << ryml::to_substr(model);
root["chars_per_sync"] << ryml::to_substr(chars_per_sync);
root["bytes_per_char"] << ryml::to_substr(bytes_per_char);
+ root["button"] << ryml::to_substr(button);
root["rows"] << ryml::to_substr(rows);
root["cols"] << ryml::to_substr(cols);
root["window_duration"] << ryml::to_substr(window_duration);
@@ -119,6 +121,7 @@ bool TranscriptionAppConfig::Deserialize(const std::filesystem::path& path) {
root.get_if("model", &c.model);
root.get_if("chars_per_sync", &c.chars_per_sync);
root.get_if("bytes_per_char", &c.bytes_per_char);
+ root.get_if("button", &c.button);
root.get_if("rows", &c.rows);
root.get_if("cols", &c.cols);
root.get_if("window_duration", &c.window_duration);
diff --git a/GUI/GUI/GUI/Config.h b/GUI/GUI/GUI/Config.h
index fe7b862..511ba01 100644
--- a/GUI/GUI/GUI/Config.h
+++ b/GUI/GUI/GUI/Config.h
@@ -41,6 +41,7 @@ public:
std::string model;
std::string chars_per_sync;
std::string bytes_per_char;
+ std::string button;
std::string rows;
std::string cols;
std::string window_duration;
diff --git a/GUI/GUI/GUI/Frame.cpp b/GUI/GUI/GUI/Frame.cpp
index 5c73bf6..e0663c5 100644
--- a/GUI/GUI/GUI/Frame.cpp
+++ b/GUI/GUI/GUI/Frame.cpp
@@ -32,6 +32,7 @@ namespace {
ID_PY_APP_MODEL,
ID_PY_APP_CHARS_PER_SYNC,
ID_PY_APP_BYTES_PER_CHAR,
+ ID_PY_APP_BUTTON,
ID_PY_APP_MODEL_PANEL,
ID_PY_APP_ENABLE_LOCAL_BEEP,
ID_PY_APP_USE_CPU,
@@ -230,6 +231,17 @@ namespace {
// Sorry international users. Optimize for English speakers, by default.
constexpr int kBytesDefault = 0;
+ const wxString kButton[] = {
+ "left joystick",
+ "left a",
+ "left b",
+ "right joystick",
+ "right a",
+ "right b",
+ };
+ const size_t kNumButtons = sizeof(kButton) / sizeof(kButton[0]);
+ constexpr int kButtonDefault = 0;
+
// Given the string value of a dropdown menu's entry, find its index. If no
// entry matches, return `default_index`.
int GetDropdownChoiceIndex(const wxString menu[],
@@ -360,6 +372,17 @@ Frame::Frame()
"characters (i.e. not English), set this to 2.");
py_app_bytes_per_char_ = py_app_bytes_per_char;
+ auto* py_app_button = new wxChoice(py_app_config_panel_pairs,
+ ID_PY_APP_BUTTON, wxDefaultPosition,
+ wxDefaultSize, kNumButtons, kButton);
+ int button_idx = GetDropdownChoiceIndex(kButton, kNumButtons, py_c.button, kButtonDefault);
+ py_app_button->SetSelection(button_idx);
+ py_app_button->SetToolTip(
+ "You will use this button in game to start and stop "
+ "transcription. Set it to a button you're not using "
+ "for anything else!");
+ py_app_button_ = py_app_button;
+
auto* py_app_rows = new wxTextCtrl(py_app_config_panel_pairs,
ID_PY_APP_ROWS, py_c.rows,
wxDefaultPosition, wxDefaultSize, /*style=*/0);
@@ -404,6 +427,9 @@ Frame::Frame()
sizer->Add(new wxStaticText(py_app_config_panel_pairs, wxID_ANY, /*label=*/"Bytes per character:"));
sizer->Add(py_app_bytes_per_char, /*proportion=*/0, /*flags=*/wxEXPAND);
+ sizer->Add(new wxStaticText(py_app_config_panel_pairs, wxID_ANY, /*label=*/"Button:"));
+ sizer->Add(py_app_button, /*proportion=*/0, /*flags=*/wxEXPAND);
+
sizer->Add(new wxStaticText(py_app_config_panel_pairs, wxID_ANY, /*label=*/"Text box rows:"));
sizer->Add(py_app_rows, /*proportion=*/0, /*flags=*/wxEXPAND);
@@ -946,6 +972,10 @@ void Frame::OnAppStart(wxCommandEvent& event) {
if (bytes_per_char_idx == wxNOT_FOUND) {
bytes_per_char_idx = kBytesDefault;
}
+ int button_idx = py_app_button_->GetSelection();
+ if (button_idx == wxNOT_FOUND) {
+ button_idx = kBytesDefault;
+ }
const bool enable_local_beep = py_app_enable_local_beep_->GetValue();
const bool use_cpu = py_app_use_cpu_->GetValue();
const bool use_builtin = py_app_use_builtin_->GetValue();
@@ -987,6 +1017,7 @@ void Frame::OnAppStart(wxCommandEvent& event) {
py_c.model = kModelChoices[which_model].ToStdString();
py_c.chars_per_sync = kCharsPerSync[chars_per_sync_idx].ToStdString();
py_c.bytes_per_char = kBytesPerChar[bytes_per_char_idx].ToStdString();
+ py_c.button = kButton[button_idx].ToStdString();
py_c.rows = std::to_string(rows);
py_c.cols = std::to_string(cols);
py_c.window_duration = std::to_string(window_duration);
diff --git a/GUI/GUI/GUI/Frame.h b/GUI/GUI/GUI/Frame.h
index c55830e..28c8f09 100644
--- a/GUI/GUI/GUI/Frame.h
+++ b/GUI/GUI/GUI/Frame.h
@@ -46,6 +46,7 @@ private:
// TODO(yum) figure out how to deduplicate these objects
wxChoice* py_app_chars_per_sync_;
wxChoice* py_app_bytes_per_char_;
+ wxChoice* py_app_button_;
wxChoice* unity_chars_per_sync_;
wxChoice* unity_bytes_per_char_;
diff --git a/GUI/GUI/GUI/PythonWrapper.cpp b/GUI/GUI/GUI/PythonWrapper.cpp
index 7571833..cad9395 100644
--- a/GUI/GUI/GUI/PythonWrapper.cpp
+++ b/GUI/GUI/GUI/PythonWrapper.cpp
@@ -154,6 +154,7 @@ wxProcess* PythonWrapper::StartApp(
"--model", config.model,
"--chars_per_sync", config.chars_per_sync,
"--bytes_per_char", config.bytes_per_char,
+ "--button", Quote(config.button),
"--enable_local_beep", config.enable_local_beep ? "1" : "0",
"--rows", config.rows,
"--cols", config.cols,
diff --git a/Scripts/steamvr.py b/Scripts/steamvr.py
index ed4150c..c231184 100644
--- a/Scripts/steamvr.py
+++ b/Scripts/steamvr.py
@@ -9,6 +9,15 @@ EVENT_NONE = 0
EVENT_RISING_EDGE = 1
EVENT_FALLING_EDGE = 2
+hands = {}
+hands["left"] = vr.TrackedControllerRole_LeftHand
+hands["right"] = vr.TrackedControllerRole_RightHand
+
+buttons = {}
+buttons["a"] = vr.k_EButton_IndexController_A
+buttons["b"] = vr.k_EButton_IndexController_B
+buttons["joystick"] = vr.k_EButton_Axis0
+
class SessionState:
def __init__(self):
self.system = vr.init(vr.VRApplication_Background)
@@ -24,11 +33,11 @@ class SessionState:
# 2 - button falling edge
def pollButtonPress(
session_state: SessionState,
- controller: vr.ETrackedControllerRole = vr.TrackedControllerRole_LeftHand,
- button: vr.EVRButtonId = vr.k_EButton_Axis0
+ hand_id: vr.ETrackedControllerRole = hands["left"],
+ button_id: vr.EVRButtonId = buttons["joystick"],
) -> int:
- lh_idx = session_state.system.getTrackedDeviceIndexForControllerRole(vr.TrackedControllerRole_LeftHand)
- #print("Left hand device idx: {}".format(lh_idx))
+ lh_idx = session_state.system.getTrackedDeviceIndexForControllerRole(hand_id)
+ #print("left hand device idx: {}".format(lh_idx))
got_state, state = session_state.system.getControllerState(lh_idx)
if not got_state:
@@ -41,13 +50,11 @@ def pollButtonPress(
# differentiate movement from clicking, we create a dead zone: if the event
# fires while the stick isn't moved far from center, we assume it's a
# click, not movement.
- dead_zone_radius = 0.5
+ dead_zone_radius = 0.7
- # This is the ID of event for the joystick being clicked.
- joy_click = vr.k_EButton_Axis0
- joy_click_mask = (1 << joy_click)
+ button_mask = (1 << button_id)
ret = EVENT_NONE
- if (state.ulButtonPressed & joy_click_mask) != 0 and\
+ if (state.ulButtonPressed & button_mask) != 0 and\
(state.rAxis[0].x**2 + state.rAxis[0].y**2 < dead_zone_radius**2):
#print("button pressed: %016x" % state.ulButtonPressed)
#for i in range(0, 5):
@@ -65,7 +72,7 @@ if __name__ == "__main__":
while True:
time.sleep(0.1)
- event = pollButtonPress(session_state)
+ event = pollButtonPress(session_state, hand_id = hands["left"], button_id = buttons["joystick"])
if event == EVENT_RISING_EDGE:
print("rising edge")
elif event == EVENT_FALLING_EDGE:
diff --git a/Scripts/transcribe.py b/Scripts/transcribe.py
index 1237334..f452b2c 100644
--- a/Scripts/transcribe.py
+++ b/Scripts/transcribe.py
@@ -317,7 +317,8 @@ def sendAudio(audio_state, use_builtin: bool):
# Pace this out
time.sleep(0.01)
-def readControllerInput(audio_state, enable_local_beep, use_builtin):
+def readControllerInput(audio_state, enable_local_beep: bool,
+ use_builtin: bool, button: str):
session = None
first = True
while session == None and audio_state.run_app == True:
@@ -334,11 +335,15 @@ def readControllerInput(audio_state, enable_local_beep, use_builtin):
osc_ctrl.indicateSpeech(audio_state.osc_state.client, False)
osc_ctrl.indicatePaging(audio_state.osc_state.client, False)
+ hand_id = steamvr.hands[button.split()[0]]
+ button_id = steamvr.buttons[button.split()[1]]
+
last_rising = time.time()
while audio_state.run_app == True:
time.sleep(0.05)
- event = steamvr.pollButtonPress(session)
+ event = steamvr.pollButtonPress(session, hand_id=hand_id,
+ button_id=button_id)
if event == steamvr.EVENT_RISING_EDGE:
last_rising = time.time()
@@ -387,7 +392,8 @@ def readControllerInput(audio_state, enable_local_beep, use_builtin):
# model should correspond to one of the Whisper models defined in
# whisper/__init__.py. Examples: tiny, base, small, medium.
def transcribeLoop(mic: str, language: str, model: str,
- enable_local_beep: bool, use_cpu: bool, use_builtin: bool):
+ enable_local_beep: bool, use_cpu: bool, use_builtin: bool,
+ button: str):
audio_state = getMicStream(mic)
audio_state.language = whisper.tokenizer.TO_LANGUAGE_CODE[language]
@@ -408,7 +414,7 @@ def transcribeLoop(mic: str, language: str, model: str,
send_audio_thd.daemon = True
send_audio_thd.start()
- controller_input_thd = threading.Thread(target = readControllerInput, args = [audio_state, enable_local_beep, use_builtin])
+ controller_input_thd = threading.Thread(target = readControllerInput, args = [audio_state, enable_local_beep, use_builtin, button])
controller_input_thd.daemon = True
controller_input_thd.start()
@@ -452,6 +458,7 @@ if __name__ == "__main__":
parser.add_argument("--window_duration_s", type=int, help="The length in seconds of the audio recording handed to the transcription algorithm")
parser.add_argument("--cpu", type=int, help="If set to 1, use CPU instead of GPU")
parser.add_argument("--use_builtin", type=int, help="If set to 1, use the text box built into the game.")
+ parser.add_argument("--button", type=str, help="The controller button used to start/stop transcription. E.g. \"left joystick\"")
args = parser.parse_args()
if not args.mic:
@@ -471,6 +478,10 @@ if __name__ == "__main__":
print("--rows and --cols required", file=sys.stderr)
sys.exit(1)
+ if not args.button:
+ print("--button required", file=sys.stderr)
+ sys.exit(1)
+
if args.window_duration_s:
config.MAX_LENGTH_S = int(args.window_duration_s)
@@ -490,5 +501,5 @@ if __name__ == "__main__":
generate_utils.config.BOARD_COLS = int(args.cols)
transcribeLoop(args.mic, args.language, args.model, args.enable_local_beep,
- args.cpu, args.use_builtin)
+ args.cpu, args.use_builtin, args.button)
generated by cgit v1.2.3 (git 2.39.1) at 2026-06-01 14:33:00 +0000