diff --git a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
index bdebb507a7b5..c7758850f750 100644
--- a/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
+++ b/src/diffusers/pipelines/stable_diffusion/pipeline_stable_unclip_img2img.py
@@ -388,7 +388,7 @@ def _encode_image(
             # what the expected dimensions of inputs should be and how we handle the encoding.
             repeat_by = num_images_per_prompt
 
-        if not image_embeds:
+        if image_embeds is None:
             if not isinstance(image, torch.Tensor):
                 image = self.feature_extractor(images=image, return_tensors="pt").pixel_values