Initial Upload

.gitignore

@@ -0,0 +1,10 @@
+# Python-generated files
+__pycache__/
+*.py[oc]
+build/
+dist/
+wheels/
+*.egg-info
+
+# Virtual environments
+.venv

.python-version

@@ -0,0 +1 @@
+3.12

main.py

@@ -0,0 +1,6 @@
+def main():
+    print("Hello from kokoro-tts!")
+
+
+if __name__ == "__main__":
+    main()

narrate.py

@@ -0,0 +1,24 @@
+import sounddevice as sd
+from kokoro import KPipeline
+import torch
+
+# Check if we are actually using your RTX 3050
+device = "cuda" if torch.cuda.is_available() else "cpu"
+print(f"Using device: {device}")
+
+# Initialize pipeline with the specific Python 3.12 environment
+pipeline = KPipeline(lang_code="a", device=device)
+
+text = """
+The RTX 3050 is now narrating this documentation. 
+Using uv and Python 3.12 ensures that we avoid the build errors 
+commonly found on Arch Linux's bleeding-edge Python 3.14.
+"""
+
+# Generating audio
+generator = pipeline(text, voice="af_heart", speed=1.1)
+
+for i, (gs, ps, audio) in enumerate(generator):
+    print(f"Playing chunk {i}...")
+    sd.play(audio, samplerate=24000)
+    sd.wait()

pyproject.toml

@@ -0,0 +1,12 @@
+[project]
+name = "kokoro-tts"
+version = "0.1.0"
+description = "Add your description here"
+readme = "README.md"
+requires-python = ">=3.12"
+dependencies = [
+    "kokoro>=0.9.4",
+    "onnxruntime-gpu>=1.23.2",
+    "sounddevice>=0.5.5",
+    "soundfile>=0.13.1",
+]

run.sh

@@ -0,0 +1,21 @@
+#!/bin/bash
+
+# Path to your project
+PROJECT_DIR="$HOME/Clones/kokoro-tts"
+
+# 1. Kill any existing narration process to act as a 'toggle' or 'reset'
+# We use pgrep -f to find the specific uv run process
+pids=$(pgrep -f "wayland_narrate.py")
+
+if [ -n "$pids" ]; then
+  kill $pids
+  # Optional: If you want it to ONLY stop when you hit the key again,
+  # uncomment the next two lines to exit here.
+  # echo "Stopped narration."
+  # exit 0
+fi
+
+# 2. Run the narrator using uv
+# we use 'nohup' and '&' so the script returns control to your desktop immediately
+cd "$PROJECT_DIR" || exit
+uv run wayland_narrate.py >/dev/null 2>&1 &

wayland_narrate.py

@@ -0,0 +1,29 @@
+import subprocess
+import sounddevice as sd
+from kokoro import KPipeline
+import torch
+import sys
+
+# 1. Fetch text from Wayland Primary Selection (highlighted text)
+try:
+    # 'wl-paste -p' gets the current mouse highlight
+    text = subprocess.check_output(["wl-paste", "-p"], text=True).strip()
+except Exception:
+    print("Nothing highlighted or wl-clipboard not installed.")
+    sys.exit(1)
+
+if not text:
+    print("Selection is empty.")
+    sys.exit(0)
+
+print(f"Narrating: {text[:50]}...")
+
+# 2. Setup Kokoro (GPU)
+device = "cuda" if torch.cuda.is_available() else "cpu"
+pipeline = KPipeline(lang_code="a", device=device)
+
+# 3. Generate and Play
+generator = pipeline(text, voice="af_heart", speed=1.1)
+for i, (gs, ps, audio) in enumerate(generator):
+    sd.play(audio, 24000)
+    sd.wait()