diff --git a/koboldcpp.py b/koboldcpp.py
index 0efa45b78..bb980c1c6 100644
--- a/koboldcpp.py
+++ b/koboldcpp.py
@@ -1228,13 +1228,16 @@ def generate(genparams, stream_flag=False):
     global showmaxctxwarning
     if max_context_length > maxctx:
         if showmaxctxwarning:
-            print(f"\n(Warning! Request max_context_length={max_context_length} exceeds allocated context size of {maxctx}. It will be reduced to fit. Consider launching with increased --contextsize to avoid errors. This message will only show once per session.)")
+            print(f"\n!!! ====== !!!\n(Warning! Request max_context_length={max_context_length} exceeds allocated context size of {maxctx}. It will be reduced to fit. Consider launching with increased --contextsize to avoid issues. This message will only show once per session.)\n!!! ====== !!!")
             showmaxctxwarning = False
         max_context_length = maxctx
-    min_remain = min(max_context_length-4, 16)
-    if max_length >= (max_context_length-min_remain):
-        max_length = max_context_length-min_remain
-        print("\nWarning: You are trying to generate with max_length near or exceeding max_context_length. Most of the context will be removed, and your outputs will not be very coherent.")
+    min_remain_hardlimit = max(min(max_context_length-4, 16),int(max_context_length*0.1))
+    min_remain_softlimit = max(min(max_context_length-4, 16),int(max_context_length*0.4))
+    if max_length >= (max_context_length-min_remain_softlimit):
+        print(f"\n!!! ====== !!!\nWarning: You are trying to generate text with max_length ({max_length}) near or exceeding max_context_length limit ({max_context_length}).\nMost of the context will be removed, and your outputs will not be very coherent.\nConsider launching with increased --contextsize to avoid issues.\n!!! ====== !!!")
+        if max_length >= (max_context_length-min_remain_hardlimit):
+            max_length = max_context_length-min_remain_hardlimit
+
 
     inputs.max_context_length = max_context_length   # this will resize the context buffer if changed
     inputs.max_length = max_length