Create data/generate_subtitles.py

data/generate_subtitles.py

@@ -0,0 +1,107 @@
+import os
+import sys
+import stable_whisper
+import json
+import argparse
+
+# --- CORRECTIF ERREUR CUDA ---
+os.environ["CUDA_VISIBLE_DEVICES"] = ""
+
+def get_style_config(theme_name):
+    """
+    Configuration des thèmes (ASS BGR). 
+    Note: On utilise 'DejaVu Sans' ou 'Liberation Sans' qui sont présents par défaut.
+    """
+    styles = {
+        "viral": {
+            "font": "DejaVu Sans Bold", 
+            "font_size": 32, 
+            "color": "00FFFF", 
+            "margin_v": 450, 
+            "outline": 4, 
+            "shadow": 2, 
+            "max_chars": 10
+        },
+        "minimal": {
+            "font": "Liberation Sans", 
+            "font_size": 22, 
+            "color": "FFFFFF", 
+            "margin_v": 350, 
+            "outline": 2, 
+            "shadow": 1, 
+            "max_chars": 15
+        },
+        "default": {
+            "font": "DejaVu Sans Bold", 
+            "font_size": 28, 
+            "color": "00FFFF", 
+            "margin_v": 400, 
+            "outline": 3, 
+            "shadow": 1.5, 
+            "max_chars": 12
+        }
+    }
+    return styles.get(theme_name, styles["default"])
+
+def run():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("audio_path")
+    parser.add_argument("output_ass")
+    parser.add_argument("text")
+    parser.add_argument("--theme", default="default")
+    parser.add_argument("--font")
+    parser.add_argument("--font_size", type=int)
+    parser.add_argument("--margin_v", type=int)
+    parser.add_argument("--color")
+    parser.add_argument("--max_chars", type=int)
+    
+    args, unknown = parser.parse_known_args()
+
+    # 1. Charger le thème de base
+    s = get_style_config(args.theme)
+
+    # 2. Override dynamique (priorité aux arguments n8n)
+    if args.font: s["font"] = args.font
+    if args.font_size: s["font_size"] = args.font_size
+    if args.margin_v: s["margin_v"] = args.margin_v
+    if args.color: s["color"] = args.color
+    if args.max_chars: s["max_chars"] = args.max_chars
+
+    try:
+        model = stable_whisper.load_model("base", device="cpu")
+        
+        # --- CHANGEMENT MAJEUR : ALIGNEMENT AU LIEU DE TRANSCRIPTION ---
+        # On force Whisper à utiliser le texte exact fourni par n8n
+        result = model.align(args.audio_path, args.text, language="fr")
+
+        # Découpage pour l'esthétique Karaoké
+        result.split_by_length(max_chars=s["max_chars"], max_words=3)
+
+        # Export au format ASS
+        result.to_ass(
+            args.output_ass,
+            word_level=True,
+            highlight_color=s["color"],
+            primary_color="FFFFFF",
+            font=s["font"],
+            font_size=s["font_size"],
+            margin_v=s["margin_v"],
+            outline=s["outline"],
+            shadow=s["shadow"],
+            alignment=2,
+            font_style='Bold=1'
+        )
+
+        print(json.dumps({
+            "status": "success", 
+            "output_file": args.output_ass, 
+            "applied_params": s,
+            "method": "alignment"
+        }))
+
+    except Exception as e:
+        print(json.dumps({"status": "error", "message": str(e)}))
+        sys.exit(1)
+
+if __name__ == "__main__":
+    run()