diff --git a/get_normal.py b/get_normal.py
index d6cbc21..d44d3d4 100644
--- a/get_normal.py
+++ b/get_normal.py
@@ -19,7 +19,7 @@ def main():
         "Stable-X/StableNormal",
         "StableNormal",
         trust_repo=True,
-        local_cache_dir="/home/lff/bigdata1/cjw/model_cache"
+        local_cache_dir="~/.cache/langscenex_models"
     )
     if not args.video_process:
         base_path = args.base_path
diff --git a/quick_start.sh b/quick_start.sh
index 8292db8..782d402 100755
--- a/quick_start.sh
+++ b/quick_start.sh
@@ -5,7 +5,7 @@ last_image=$2
 
 sam_model_path="./sam_vit_h_4b8939.pth"
 sam2_model_path="./sam2_hiera_large.pt"
-model_path="./CogVideo-X"
+model_path="./CogVideoX-ft"
 
 mkdir -p demo/rgb
 cp $first_image demo/rgb/0001.png
@@ -22,7 +22,7 @@ python auto-seg/auto-mask-align.py \
 # get normal maps
 python get_normal.py --base_path demo
 
-RGB video interpolation
+# RGB video interpolation
 python video_inference.py \
     --model_path $model_path \
     --output_dir demo/video/rgb \
@@ -58,4 +58,4 @@ python entry_point.py \
     gaussian.opt.optim_pose=False \
     pipeline.skip_video_process=False \
     pipeline.skip_pose_estimate=False \
-    pipeline.skip_lang_feature_extraction=False
\ No newline at end of file
+    pipeline.skip_lang_feature_extraction=False