diff --git a/invokeai/app/invocations/latent.py b/invokeai/app/invocations/latent.py
index e9b1425464..fad875cdbb 100644
--- a/invokeai/app/invocations/latent.py
+++ b/invokeai/app/invocations/latent.py
@@ -286,6 +286,10 @@ class TextToLatentsInvocation(BaseInvocation):
             )
         )
 
+        latents_shape = noise.shape
+        control_height_resize = latents_shape[2] * 8
+        control_width_resize = latents_shape[3] * 8
+
         # copied from old backend/txt2img.py
         # FIXME: still need to test with different widths, heights, devices, dtypes
         #        and add in batch_size, num_images_per_prompt?
@@ -295,10 +299,8 @@ class TextToLatentsInvocation(BaseInvocation):
                     image=control_image,
                     # do_classifier_free_guidance=do_classifier_free_guidance,
                     do_classifier_free_guidance=True,
-                    # width=width,
-                    # height=height,
-                    width=512,
-                    height=512,
+                    width=control_width_resize,
+                    height=control_height_resize,
                     # batch_size=batch_size * num_images_per_prompt,
                     # num_images_per_prompt=num_images_per_prompt,
                     device=control_model.device,
@@ -311,10 +313,8 @@ class TextToLatentsInvocation(BaseInvocation):
                         image=image_,
                         # do_classifier_free_guidance=do_classifier_free_guidance,
                         do_classifier_free_guidance=True,
-                        # width=width,
-                        # height=height,
-                        width=512,
-                        height=512,
+                        width=control_width_resize,
+                        height=control_height_resize,
                         # batch_size=batch_size * num_images_per_prompt,
                         # num_images_per_prompt=num_images_per_prompt,
                         device=control_model.device,
@@ -323,10 +323,7 @@ class TextToLatentsInvocation(BaseInvocation):
                     images.append(image_)
                 control_image = images
 
-
-
         # TODO: Verify the noise is the right size
-
         result_latents, result_attention_map_saver = model.latents_from_embeddings(
             latents=torch.zeros_like(noise, dtype=torch_dtype(model.device)),
             noise=noise,
diff --git a/invokeai/backend/stable_diffusion/diffusers_pipeline.py b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
index 7d88b0e07a..e24f11fd3d 100644
--- a/invokeai/backend/stable_diffusion/diffusers_pipeline.py
+++ b/invokeai/backend/stable_diffusion/diffusers_pipeline.py
@@ -994,14 +994,16 @@ class StableDiffusionGeneratorPipeline(StableDiffusionPipeline):
         self,
         image,
         # FIXME: need to fix hardwiring of width and height, change to basing on latents dimensions?
-        width=512,
-        height=512,
+        # latents,
+        width=512,  # should be 8 * latent.shape[3]
+        height=512, # should be 8 * latent height[2]
         batch_size=1,
         num_images_per_prompt=1,
         device="cuda",
         dtype=torch.float16,
         do_classifier_free_guidance=True,
     ):
+
         if not isinstance(image, torch.Tensor):
             if isinstance(image, PIL.Image.Image):
                 image = [image]