WIP animate mouth scale based on precalculated voiceline amplitude timelines

3d-generation-pipeline/generate_voiceline_amplitude_curves.py

@@ -0,0 +1,52 @@
+import os
+import librosa
+import numpy as np
+
+# === CONFIG ===
+FRAME_DURATION = 0.02  # 20 ms windows
+GAIN = 1.0             # multiply RMS values (set to e.g. 30.0 if you want larger values)
+EXTENSION = ".txt"     # output file extension
+
+
+def process_wav(filepath):
+    print(f"Processing: {filepath}")
+
+    # Load audio
+    audio, sr = librosa.load(filepath, mono=True)
+
+    # Frame size in samples
+    frame_len = int(FRAME_DURATION * sr)
+    hop_len = frame_len
+
+    # Compute RMS
+    rms = librosa.feature.rms(
+        y=audio,
+        frame_length=frame_len,
+        hop_length=hop_len
+    )[0]
+
+    # Apply optional gain
+    rms = rms * GAIN
+
+    # Save to .txt
+    out_path = os.path.splitext(filepath)[0] + EXTENSION
+    np.savetxt(out_path, rms, fmt="%.8f")
+
+    print(f"Saved RMS → {out_path}")
+
+
+def process_folder(folder_path):
+    print(f"Scanning folder: {folder_path}")
+
+    for filename in os.listdir(folder_path):
+        if filename.lower().endswith(".wav"):
+            filepath = os.path.join(folder_path, filename)
+            process_wav(filepath)
+
+    print("Done!")
+
+
+# === Run script ===
+if __name__ == "__main__":
+    folder = input("Enter folder path: ").strip()
+    process_folder(folder)