Update app.py

app.py

@@ -394,7 +394,7 @@ def generate_final(image, frames, seed):
         width=TARGET_W,
         cfg_scale=5.0,
         num_frames=29,
-        num_inference_steps=50,
+        num_inference_steps=28,
         seed=int(seed),
         tiled=True
     )
@@ -425,7 +425,7 @@ with gr.Blocks() as demo:
                 <li style="margin-bottom: 4px;"><b>Scale & Adjustment / 尺度与调整：</b>The motion scale is normalized by VGGT, and the final point cloud is provided to help adjust motion parameters.<br>
                 所有运动数值由 VGGT 统一尺度建模，最终提供的点云结果可用于辅助调整相机运动参数。</li>
                 
-                <li><b>First Run / 首次运行：</b>Please note that the first execution will take slightly longer as the models are being loaded into the GPU. <br>
+                <li><b>Tips / 提示：</b>Default inference steps: 28 (Speed & Quality balanced). Run locally with higher steps for better results. <br>
                 首次运行需要将模型权重加载到显存，耗时会稍微久一点，请耐心等待。</li>
             </ul>
         </div>""")