diff --git a/mlx_video/generate.py b/mlx_video/generate.py
index e39dfeb..0c9e562 100644
--- a/mlx_video/generate.py
+++ b/mlx_video/generate.py
@@ -265,7 +265,7 @@ def generate_video(
 
     vae_decoder = load_vae_decoder(
         str(model_path / 'ltx-2-19b-distilled.safetensors'),
-        timestep_conditioning=True
+        timestep_conditioning=None  # Auto-detect from model metadata
     )
 
     latents = upsample_latents(latents, upsampler, vae_decoder.latents_mean, vae_decoder.latents_std)
diff --git a/mlx_video/generate_av.py b/mlx_video/generate_av.py
index b4e488c..7551c51 100644
--- a/mlx_video/generate_av.py
+++ b/mlx_video/generate_av.py
@@ -427,7 +427,7 @@ def generate_video_with_audio(
 
     vae_decoder = load_vae_decoder(
         str(model_path / 'ltx-2-19b-distilled.safetensors'),
-        timestep_conditioning=True
+        timestep_conditioning=None  # Auto-detect from model metadata
     )
 
     video_latents = upsample_latents(video_latents, upsampler, vae_decoder.latents_mean, vae_decoder.latents_std)