diff --git a/src/h3_blackwell_runtime/sampler.py b/src/h3_blackwell_runtime/sampler.py
index ec2537d..598656b 100644
--- a/src/h3_blackwell_runtime/sampler.py
+++ b/src/h3_blackwell_runtime/sampler.py
@@ -61,13 +61,13 @@ def sample_video_res_multistep(model, packer: H3PromptPacker, text: torch.Tensor
         previous_index = index - 1
         sigma_down = sigmas[index]
         sigma_audio = _audio_sigma(sigma)
-        native_audio = audio_carried * (sigma_audio / sigma)
+        carry = sigma_audio / sigma
+        native_audio = audio_carried.to(torch.bfloat16) * carry
         hidden, times, segments, positions, video_segment, audio_segment = packer(text, video, native_audio, float(sigma))
         raw_video, raw_audio = model(hidden, times, positions, segments, video_segment, audio_segment)
         raw_video = raw_video.to(torch.bfloat16).float()
         raw_audio = raw_audio.to(torch.bfloat16)
         velocity_video = -unpatchify_video(raw_video, video.shape[2], video.shape[-2], video.shape[-1])
-        carry = sigma_audio / sigma
         velocity_audio = (
             (1.0 - 4.0) * (audio_carried.to(torch.bfloat16) * carry.to(torch.bfloat16))
             + (1.0 + 3.0 * sigma_audio).to(torch.bfloat16) * (-_unpack_audio(raw_audio))