diff --git a/data_chain/parser/tools/ocr.py b/data_chain/parser/tools/ocr.py
index e49c465637cfdbd2f7bb974a2890c8099abfdc64..1d529edad38bd2fbf7e0f0a5ac6b32b6437029aa 100644
--- a/data_chain/parser/tools/ocr.py
+++ b/data_chain/parser/tools/ocr.py
@@ -100,9 +100,12 @@ class BaseOCR:
                 return ''
             pre_part_description = ""
             ocr_result_parts = await self.cut_ocr_result_in_part(ocr_result, self.max_tokens // 5*2)
+            image_related_text = split_tools.get_k_tokens_words_from_content(image_related_text, self.max_tokens // 5*2)
             user_call = '请详细输出图片的摘要，不要输出其他内容'
             for part in ocr_result_parts:
                 pre_part_description_cp = pre_part_description
+                pre_part_description = split_tools.get_k_tokens_words_from_content(
+                    pre_part_description, self.max_tokens // 5)
                 try:
                     prompt = prompt_template.format(
                         image_related_text=image_related_text,
diff --git a/data_chain/parser/tools/split.py b/data_chain/parser/tools/split.py
index 2de56ec09204bcb16975d467ff4c74eebdcd6d1e..d4c5d9a03d68aa79432c34c29894bc4ec0976c0b 100644
--- a/data_chain/parser/tools/split.py
+++ b/data_chain/parser/tools/split.py
@@ -13,6 +13,24 @@ class SplitTools:
             logging.error(f"Get tokens failed due to: {e}")
             return 0
 
+    def get_k_tokens_words_from_content(self, content: str, k: int) -> list:
+        try:
+            if (self.get_tokens(content) <= k):
+                return content
+            l = 0
+            r = len(content)
+            while l+1 < r:
+                mid = (l+r)//2
+                if (self.get_tokens(content[:mid]) <= k):
+                    l = mid
+                else:
+                    r = mid
+            return content[:l]
+        except Exception as e:
+            err = f"[TokenTool] 获取k个token的词失败 {e}"
+            logging.exception("[TokenTool] %s", err)
+        return ""
+
     def split_words(self, text):
         return list(jieba.cut(str(text)))