Update app.py

app.py

@@ -1163,17 +1163,17 @@ def img_to_image(image, prompt, neg_prompt, strength, steps, guidance_scale,
                 ui_strength = strength  
     
                 # 1. Basierend auf der UI-Stärke (strength) berechnen
-                adj_strength = ui_strength * 0.8
-                controlnet_strength = 0.6 - (ui_strength * 0.4)
-                pose_ratio = 0.75 - (ui_strength * 0.1)
-                canny_ratio = 0.25 + (ui_strength * 0.1)
+                adj_strength = 0.15 + (ui_strength * 0.75)
+                controlnet_strength = 0.8 - (ui_strength * 0.4)
+                pose_ratio = 0.8 - (ui_strength * 0.3)
+                canny_ratio = 0.2 + (ui_strength * 0.3)
          
                 
                 # 2. Werte auf sinnvolle Bereiche begrenzen (Clipping)
-                adj_strength = max(0.6, min(adj_strength, 0.9))
-                controlnet_strength = max(0.15, min(controlnet_strength, 0.6))
-                pose_ratio = max(0.6, min(pose_ratio, 0.8))
-                canny_ratio = max(0.2, min(canny_ratio, 0.4))
+                adj_strength = max(0.15, min(adj_strength, 0.85))
+                controlnet_strength = max(0.3, min(controlnet_strength, 0.8))
+                pose_ratio = max(0.5, min(pose_ratio, 0.8))
+                canny_ratio = max(0.2, min(canny_ratio, 0.5))
 
                 conditioning_scale = [
                     controlnet_strength * pose_ratio,  # Depth-Gewichtung
@@ -1183,6 +1183,7 @@ def img_to_image(image, prompt, neg_prompt, strength, steps, guidance_scale,
                 print(f"👤 Humanoid → Humanoid (UI-Stärke: {ui_strength})")
                 print(f"   adj_strength: {adj_strength:.2f}, controlnet: {controlnet_strength:.2f}")
                 print(f"   Verhältnis: Pose {pose_ratio*100:.0f}% : Canny {canny_ratio*100:.0f}%")
+                print(f"   Scale: [{conditioning_scale[0]:.3f}, {conditioning_scale[1]:.3f}]")