diff --git a/README.md b/README.md
index 88e5221..71d30e1 100644
--- a/README.md
+++ b/README.md
@@ -13,7 +13,7 @@ The sp file format is shown below.
 [struct_len] + n*[[type][data_len][data]]
 ```
 
-1. `type` is an one-byte number indicating the type length(2^type^bytes) of this struct item.
+1. `type` is a length-variable LE number(less than 2^57^) indicating the aligned type length of each struct item.
 2. `*_len` is a length-variable LE number(less than 2^57^) indicating the length of `*`.
 3. `data` contains `data_len` bytes of the data.
 
@@ -21,18 +21,11 @@ The sp file format is shown below.
 You can read `test.c` to find out the detailed usage.
 
 ## Save a struct
-### 1. Create items type info array
+### 1. Create items type length info array
 
-This array has type `uint8_t`, which indicates the length of each struct item. Note that the length of each item must be 2^n^.
-
-### 2. Align items type info array
-
-You can use function `align_struct` to align each item to the proper size.
-
-If you are sure that your struct has been aligned, you won't need to run `align_struct`.
-
-### 3. Call set_pb to save
+This array has type `uint64_t`, which indicates the aligned length of each struct item. You should call `align_struct` to get that array.
 
+### 2. Call set_pb to save
 
 ## Read into a struct
 
diff --git a/README_ZH.md b/README_ZH.md
index f99fc15..ce29f68 100644
--- a/README_ZH.md
+++ b/README_ZH.md
@@ -11,7 +11,7 @@ Google ProtoBuf的简化变体。
 [struct_len] + n*[[type][data_len][data]]
 ```
 
-1. `type` 指明该项的类型，亦即其在结构体中实际占用的空间的对数。
+1. `type` 是一个小于 2^57^ 的变长数字，指明该项的类型，亦即其在结构体中对齐后实际占用的空间。
 2. `*_len` 是一个小于 2^57^ 的变长数字。
 3. `data` 是 `data_len` 长度的数据。
 
@@ -19,15 +19,11 @@ Google ProtoBuf的简化变体。
 比较简单，详见`test.c`。
 
 ## 保存结构体
-### 1. 创建 items_type 数组
+### 1. 调用 align_struct 创建 items_len 数组
 
-是结构体所有项的`type`组成的数组。
+该数组由结构体所有项的`type`组成。
 
-### 2. 类型对齐
-
-接下来需要用`align_struct`进行对齐。如果您已经将数组对齐或结构体本身是对齐的，则无需此步骤。
-
-### 3. 调用 set_pb 保存
+### 2. 调用 set_pb 保存
 
 
 ## 读入结构体