diff --git a/configs/animatediff/README.md b/configs/animatediff/README.md
index 1432999fe..a1337427a 100644
--- a/configs/animatediff/README.md
+++ b/configs/animatediff/README.md
@@ -35,6 +35,12 @@ This model has several weights including vae, unet and clip. You should download
 
 Running the following codes, you can get a text-generated image.
 
+### Reccomendation
+
+It's highly recommended to install [xformers](https://github.com/facebookresearch/xformers). It would save about 20G memory for 512\*512 resolution generation.
+
+### Steps
+
 1. Download [ToonYou](https://civitai.com/api/download/models/78775) and MotionModule checkpoint
 
 ```bash
@@ -52,11 +58,11 @@ wget https://civitai.com/api/download/models/78775 -P DreamBooth_LoRA/ --content
 ```python
     models_path = {Your Checkpoints Path}
     motion_module_cfg=dict(
-        path={Your MotionModule path}
+        path={Your MotionModule Path}
     ),
     dream_booth_lora_cfg=dict(
         type='ToonYou',
-        path={Your Dreambooth_Lora path},
+        path={Your Dreambooth_Lora Path},
         steps=25,
         guidance_scale=7.5)
 ```
@@ -106,7 +112,7 @@ time_str = datetime.datetime.now().strftime("%Y-%m-%dT%H-%M-%S")
 savedir = f"samples/{Path(cfg.model['dream_booth_lora_cfg']['type']).stem}-{time_str}"
 os.makedirs(savedir)
 for prompt_idx, (prompt, n_prompt, random_seed) in enumerate(zip(prompts, negative_prompts, random_seeds)):
-    output_dict = animatediff.infer(prompt,negative_prompt=n_prompt, video_length=16, height=512, width=512, seed=random_seed,num_inference_steps=cfg.model['dream_booth_lora_cfg']['steps'])
+    output_dict = animatediff.infer(prompt,negative_prompt=n_prompt, video_length=16, height=256, width=256, seed=random_seed,num_inference_steps=cfg.model['dream_booth_lora_cfg']['steps'])
     sample = output_dict['samples']
     prompt = "-".join((prompt.replace("/", "").split(" ")[:10]))
     save_videos_grid(sample, f"{savedir}/sample/{sample_idx}-{prompt}.gif")
diff --git a/configs/animatediff/README_zh-CN.md b/configs/animatediff/README_zh-CN.md
index 239e52dd3..9d2165e75 100644
--- a/configs/animatediff/README_zh-CN.md
+++ b/configs/animatediff/README_zh-CN.md
@@ -33,10 +33,14 @@
 |    [RealisticVision](./animatediff_RealisticVision.py)    |              [model](https://civitai.com/api/download/models/29460)               |
 | [RealisticVision_v2](./animatediff_RealisticVision_v2.py) | [model](https://huggingface.co/guoyww/animatediff/resolve/main/mm_sd_v15_v2.ckpt) |
 
-## Quick Start
+## 快速开始
 
 运行以下代码，你可以使用AnimateDiff通过文本生成视频。
 
+#### 小建议
+
+强烈推荐安装 [xformers](https://github.com/facebookresearch/xformers)，512\*512分辨率可以节省约20G显存。
+
 1. 下载 [ToonYou](https://civitai.com/api/download/models/78775) 和 MotionModule 权重
 
 ```bash
@@ -54,11 +58,11 @@ wget https://civitai.com/api/download/models/78775 -P DreamBooth_LoRA/ --content
 ```python
     models_path = {Your Checkpoints Path}
     motion_module_cfg=dict(
-        path={Your MotionModule path}
+        path={Your MotionModule Path}
     ),
     dream_booth_lora_cfg=dict(
         type='ToonYou',
-        path={Your Dreambooth_Lora path},
+        path={Your Dreambooth_Lora Path},
         steps=25,
         guidance_scale=7.5)
 ```
@@ -107,7 +111,7 @@ time_str = datetime.datetime.now().strftime("%Y-%m-%dT%H-%M-%S")
 savedir = f"samples/{Path(cfg.model['dream_booth_lora_cfg']['type']).stem}-{time_str}"
 os.makedirs(savedir)
 for prompt_idx, (prompt, n_prompt, random_seed) in enumerate(zip(prompts, negative_prompts, random_seeds)):
-    output_dict = animatediff.infer(prompt,negative_prompt=n_prompt, video_length=16, height=512, width=512, seed=random_seed,num_inference_steps=cfg.model['dream_booth_lora_cfg']['steps'])
+    output_dict = animatediff.infer(prompt,negative_prompt=n_prompt, video_length=16, height=256, width=256, seed=random_seed,num_inference_steps=cfg.model['dream_booth_lora_cfg']['steps'])
     sample = output_dict['samples']
     prompt = "-".join((prompt.replace("/", "").split(" ")[:10]))
     save_videos_grid(sample, f"{savedir}/sample/{sample_idx}-{prompt}.gif")
@@ -189,7 +193,7 @@ prompt:
 
 ```
 
-## Citation
+## 引用
 
 ```bibtex
 @article{guo2023animatediff,
diff --git a/mmagic/models/editors/animatediff/animatediff_utils.py b/mmagic/models/editors/animatediff/animatediff_utils.py
index 1d513d284..44d8923f8 100644
--- a/mmagic/models/editors/animatediff/animatediff_utils.py
+++ b/mmagic/models/editors/animatediff/animatediff_utils.py
@@ -1096,4 +1096,4 @@ def save_videos_grid(videos: torch.Tensor,
         outputs.append(x)
 
     os.makedirs(os.path.dirname(path), exist_ok=True)
-    imageio.mimsave(path, outputs, fps=fps)
+    imageio.mimsave(path, outputs, duration=1000 * 1 / fps)
diff --git a/mmagic/models/editors/animatediff/attention_3d.py b/mmagic/models/editors/animatediff/attention_3d.py
index 58fa2f7b0..de8464ddd 100644
--- a/mmagic/models/editors/animatediff/attention_3d.py
+++ b/mmagic/models/editors/animatediff/attention_3d.py
@@ -467,7 +467,10 @@ def __init__(
         # You can set slice_size with `set_attention_slice`
         self.sliceable_head_dim = heads
         self._slice_size = None
-        self._use_memory_efficient_attention_xformers = True
+        if xformers is not None:
+            self._use_memory_efficient_attention_xformers = True
+        else:
+            self._use_memory_efficient_attention_xformers = False
         self.added_kv_proj_dim = added_kv_proj_dim
 
         if norm_num_groups is not None: