diff --git a/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py b/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
index 30a12c2ba41f..1fe958816ea4 100644
--- a/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
+++ b/src/diffusers/pipelines/aura_flow/pipeline_aura_flow.py
@@ -607,6 +607,9 @@ def __call__(
         )
 
         # 6. Denoising loop
+        # We set the index here to remove DtoH sync, helpful especially during compilation.
+        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
+        self.scheduler.set_begin_index(0)
         num_warmup_steps = max(len(timesteps) - num_inference_steps * self.scheduler.order, 0)
         self._num_timesteps = len(timesteps)
         with self.progress_bar(total=num_inference_steps) as progress_bar:
diff --git a/src/diffusers/pipelines/chroma/pipeline_chroma.py b/src/diffusers/pipelines/chroma/pipeline_chroma.py
index 3d91f43b07cf..c16cd8155a70 100644
--- a/src/diffusers/pipelines/chroma/pipeline_chroma.py
+++ b/src/diffusers/pipelines/chroma/pipeline_chroma.py
@@ -840,6 +840,9 @@ def __call__(
             )
 
         # 6. Denoising loop
+        # We set the index here to remove DtoH sync, helpful especially during compilation.
+        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
+        self.scheduler.set_begin_index(0)
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
                 if self.interrupt:
diff --git a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
index 5e785c38891c..9e7c198c19cc 100644
--- a/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
+++ b/src/diffusers/pipelines/hunyuan_video/pipeline_hunyuan_video.py
@@ -659,6 +659,9 @@ def __call__(
         guidance = torch.tensor([guidance_scale] * latents.shape[0], dtype=transformer_dtype, device=device) * 1000.0
 
         # 7. Denoising loop
+        # We set the index here to remove DtoH sync, helpful especially during compilation.
+        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
+        self.scheduler.set_begin_index(0)
         num_warmup_steps = len(timesteps) - num_inference_steps * self.scheduler.order
         self._num_timesteps = len(timesteps)
 
diff --git a/src/diffusers/pipelines/lumina2/pipeline_lumina2.py b/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
index 2129bfeb8fbf..eb376f9bd9cc 100644
--- a/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
+++ b/src/diffusers/pipelines/lumina2/pipeline_lumina2.py
@@ -663,6 +663,9 @@ def __call__(
         self._num_timesteps = len(timesteps)
 
         # 6. Denoising loop
+        # We set the index here to remove DtoH sync, helpful especially during compilation.
+        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
+        self.scheduler.set_begin_index(0)
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
                 # compute whether apply classifier-free truncation on this timestep
diff --git a/src/diffusers/pipelines/mochi/pipeline_mochi.py b/src/diffusers/pipelines/mochi/pipeline_mochi.py
index 7108d4ff774f..c146d2d1e564 100644
--- a/src/diffusers/pipelines/mochi/pipeline_mochi.py
+++ b/src/diffusers/pipelines/mochi/pipeline_mochi.py
@@ -633,6 +633,9 @@ def __call__(
         self._num_timesteps = len(timesteps)
 
         # 6. Denoising loop
+        # We set the index here to remove DtoH sync, helpful especially during compilation.
+        # Check out more details here: https://github.com/huggingface/diffusers/pull/11696
+        self.scheduler.set_begin_index(0)
         with self.progress_bar(total=num_inference_steps) as progress_bar:
             for i, t in enumerate(timesteps):
                 if self.interrupt: