The most powerful and modular diffusion model GUI, api and backend with a graph/nodes interface. The fastest local inference engine in the world.
Negative audio in s2v should be zeros. (#9578)
comfyanonymous31a37686d02a
diff --git a/comfy_extras/nodes_wan.py b/comfy_extras/nodes_wan.pyindex 89ff74d85..312260f00 100644--- a/comfy_extras/nodes_wan.py+++ b/comfy_extras/nodes_wan.py@@ -920,7 +920,7 @@ class WanSoundImageToVideo(io.ComfyNode):audio_embed_bucket = audio_embed_bucket.permute(0, 2, 3, 1)positive = node_helpers.conditioning_set_values(positive, {"audio_embed": audio_embed_bucket})- negative = node_helpers.conditioning_set_values(negative, {"audio_embed": audio_embed_bucket})+ negative = node_helpers.conditioning_set_values(negative, {"audio_embed": audio_embed_bucket * 0.0})if ref_image is not None:ref_image = comfy.utils.common_upscale(ref_image[:1].movedim(-1, 1), width, height, "bilinear", "center").movedim(1, -1)