diff --git a/hyvideo/modules/models.py b/hyvideo/modules/models.py
index 5ca086d..b0d4d97 100644
--- a/hyvideo/modules/models.py
+++ b/hyvideo/modules/models.py
@@ -597,7 +597,8 @@ class HYVideoDiffusionTransformer(ModelMixin, ConfigMixin):
         vec = self.time_in(t)
 
         # text modulation
-        vec = vec + self.vector_in(text_states_2)
+        if text_states_2 is not None:
+            vec = vec + self.vector_in(text_states_2)
 
         # guidance modulation
         if self.guidance_embed: