Switch PTT to evdev for reliable key state (no repeat spam)

pynput on X11 sees key repeats as real press/release pairs, causing
rapid-fire toggling. evdev gives event.value=2 for repeats which
we simply ignore. Only real press (1) and release (0) are handled.

Requires sudo or 'input' group membership.

https://claude.ai/code/session_01XKYC1basxdwtHy71tky7xm

openwhispr-ptt.py

@@ -2,52 +2,39 @@
 """
 openwhispr-ptt.py — Push-to-talk wrapper for OpenWhispr
 
-Turns Ctrl+Alt (hold) into Ctrl+` toggle pairs:
+Turns Ctrl+Alt (hold) into Ctrl+` toggle pairs using raw evdev input.
-  - Press Ctrl+Alt  → sends Ctrl+` to START dictation
+No key repeat issues — reads physical key state directly.
-  - Release Ctrl+Alt → sends Ctrl+` to STOP dictation
 
-Requires: pynput, xdotool
+Requires: evdev, xdotool
-Usage: python3 openwhispr-ptt.py
+May need to run as root or add user to 'input' group.
+Usage: sudo python3 openwhispr-ptt.py
 """
 
 import subprocess
-import threading
-import time
 import signal
 import sys
 
 try:
-    from pynput import keyboard
+    import evdev
+    from evdev import ecodes
 except ImportError:
-    print("Error: pynput not installed.")
+    print("Error: evdev not installed.")
-    print("Install with: pip3 install pynput")
+    print("Install with: pip3 install evdev --break-system-packages")
     sys.exit(1)
 
-# ── Configuration ────────────────────────────────────────────────────────────
-
-# The key combo YOU hold to talk
-PTT_MODIFIER = keyboard.Key.ctrl_l
-PTT_KEY = keyboard.Key.alt_l
-
-# Also accept right-hand variants
-PTT_MODIFIER_ALT = keyboard.Key.ctrl_r
-PTT_KEY_ALT = keyboard.Key.alt_r
-
-# The key combo OpenWhispr uses (Ctrl+`)
-# We simulate this via xdotool for reliability with Electron apps
 WHISPR_HOTKEY = "ctrl+grave"
 
-# ── State ────────────────────────────────────────────────────────────────────
+KEY_LEFTCTRL = ecodes.KEY_LEFTCTRL
+KEY_RIGHTCTRL = ecodes.KEY_RIGHTCTRL
+KEY_LEFTALT = ecodes.KEY_LEFTALT
+KEY_RIGHTALT = ecodes.KEY_RIGHTALT
 
-is_recording = False
-press_time = 0
 ctrl_held = False
-super_held = False
+alt_held = False
-lock = threading.Lock()
+is_recording = False
 
 
 def send_whispr_toggle():
-    """Send Ctrl+` to OpenWhispr via xdotool."""
     try:
         subprocess.run(
             ["xdotool", "key", WHISPR_HOTKEY],
@@ -55,66 +42,75 @@ def send_whispr_toggle():
             capture_output=True,
         )
     except FileNotFoundError:
-        print("Error: xdotool not installed. Install with: sudo apt install xdotool")
+        print("Error: xdotool not installed.")
         sys.exit(1)
     except subprocess.TimeoutExpired:
         pass
 
 
-def on_press(key):
+def find_keyboard():
-    global is_recording, press_time, ctrl_held, super_held
+    """Find the keyboard device."""
-
+    devices = [evdev.InputDevice(path) for path in evdev.list_devices()]
-    with lock:
+    for dev in devices:
-        # Track modifier state — ignore key repeat (already True)
+        caps = dev.capabilities(verbose=False)
-        if key in (PTT_MODIFIER, PTT_MODIFIER_ALT):
+        if ecodes.EV_KEY in caps:
-            ctrl_held = True
+            keys = caps[ecodes.EV_KEY]
-        elif key in (PTT_KEY, PTT_KEY_ALT):
+            if ecodes.KEY_A in keys and ecodes.KEY_LEFTCTRL in keys:
-            super_held = True
+                return dev
-
+    return None
-        # Start recording when both held and not already recording
-        if ctrl_held and super_held and not is_recording:
-            press_time = time.time()
-            is_recording = True
-            send_whispr_toggle()  # START
-            print("\033[32m● Recording...\033[0m", flush=True)
-
-
-def on_release(key):
-    global is_recording, press_time, ctrl_held, super_held
-
-    with lock:
-        released_ptt = False
-
-        if key in (PTT_MODIFIER, PTT_MODIFIER_ALT):
-            ctrl_held = False
-            released_ptt = True
-        elif key in (PTT_KEY, PTT_KEY_ALT):
-            super_held = False
-            released_ptt = True
-
-        # Stop recording when either key is truly released
-        if is_recording and released_ptt:
-            elapsed_ms = (time.time() - press_time) * 1000
-            send_whispr_toggle()  # STOP
-            is_recording = False
-            print(f"\033[0m○ Stopped ({int(elapsed_ms)}ms)", flush=True)
 
 
 def main():
-    print("═══════════════════════════════════════════")
+    global ctrl_held, alt_held, is_recording
-    print("  OpenWhispr Push-to-Talk")
+
-    print("═══════════════════════════════════════════")
+    print("=" * 43)
-    print("  Hold:    Ctrl + Alt  →  start dictation")
+    print("  OpenWhispr Push-to-Talk (evdev)")
-    print("  Release: Ctrl + Alt  →  stop dictation")
+    print("=" * 43)
+    print("  Hold:    Ctrl + Alt  ->  start dictation")
+    print("  Release: Ctrl + Alt  ->  stop dictation")
     print("  Quit:    Ctrl + C")
-    print("═══════════════════════════════════════════")
+    print("=" * 43)
+
+    kb = find_keyboard()
+    if kb is None:
+        print("")
+        print("Error: No keyboard found.")
+        print("Try: sudo python3 ~/openwhispr-ptt.py")
+        print("Or: sudo usermod -aG input $USER  (then log out/in)")
+        sys.exit(1)
+
+    print("  Keyboard: " + kb.name)
+    print("=" * 43)
     print("Listening...\n")
 
-    # Handle Ctrl+C gracefully
     signal.signal(signal.SIGINT, lambda *_: sys.exit(0))
 
-    with keyboard.Listener(on_press=on_press, on_release=on_release) as listener:
+    for event in kb.read_loop():
-        listener.join()
+        if event.type != ecodes.EV_KEY:
+            continue
+
+        # value: 0=release, 1=press, 2=repeat — IGNORE repeats
+        if event.value == 2:
+            continue
+
+        pressed = (event.value == 1)
+
+        if event.code in (KEY_LEFTCTRL, KEY_RIGHTCTRL):
+            ctrl_held = pressed
+        elif event.code in (KEY_LEFTALT, KEY_RIGHTALT):
+            alt_held = pressed
+
+        # Start on both pressed
+        if ctrl_held and alt_held and not is_recording:
+            is_recording = True
+            send_whispr_toggle()
+            print("\033[32m● Recording...\033[0m", flush=True)
+
+        # Stop when either released
+        if is_recording and not (ctrl_held and alt_held):
+            send_whispr_toggle()
+            is_recording = False
+            print("\033[0m○ Stopped", flush=True)
 
 
 if __name__ == "__main__":