diff --git a/src/diffusers/pipelines/flux2/pipeline_flux2.py b/src/diffusers/pipelines/flux2/pipeline_flux2.py
index b54a43dd89a5..7e7c027736d7 100644
--- a/src/diffusers/pipelines/flux2/pipeline_flux2.py
+++ b/src/diffusers/pipelines/flux2/pipeline_flux2.py
@@ -803,6 +803,11 @@ def __call__(
                 Pre-generated noisy latents, sampled from a Gaussian distribution, to be used as inputs for image
                 generation. Can be used to tweak the same generation with different prompts. If not provided, a latents
                 tensor will be generated by sampling using the supplied random `generator`.
+                
+                For this pipeline, the expected latent shape is `(batch_size, 128, height // 16, width // 16)` when using
+                the default `AutoencoderKLFlux2` VAE shipped with FLUX.2. Here `height` and `width` correspond to the
+                requested output resolution. Passing latents with a different shape will lead to shape mismatches inside
+                the VAE and transformer.
             prompt_embeds (`torch.Tensor`, *optional*):
                 Pre-generated text embeddings. Can be used to easily tweak text inputs, *e.g.* prompt weighting. If not
                 provided, text embeddings will be generated from `prompt` input argument.