From df6516a71645b765ab446612fd36adb7c79267e8 Mon Sep 17 00:00:00 2001 From: Samu Tamminen Date: Sat, 10 Jan 2026 01:51:04 +0200 Subject: [PATCH] Align HunyuanVideoConditionEmbedding with CombinedTimestepGuidanceTextProjEmbeddings (#12316) conditioning additions inline with CombinedTimestepGuidanceTextProjEmbeddings Co-authored-by: Samu Tamminen Co-authored-by: YiYi Xu --- .../models/transformers/transformer_hunyuan_video.py | 6 +++--- 1 file changed, 3 insertions(+), 3 deletions(-) diff --git a/src/diffusers/models/transformers/transformer_hunyuan_video.py b/src/diffusers/models/transformers/transformer_hunyuan_video.py index fb0ce1a30f..4f0775ac9f 100644 --- a/src/diffusers/models/transformers/transformer_hunyuan_video.py +++ b/src/diffusers/models/transformers/transformer_hunyuan_video.py @@ -312,7 +312,6 @@ class HunyuanVideoConditionEmbedding(nn.Module): timesteps_proj = self.time_proj(timestep) timesteps_emb = self.timestep_embedder(timesteps_proj.to(dtype=pooled_projection.dtype)) # (N, D) pooled_projections = self.text_embedder(pooled_projection) - conditioning = timesteps_emb + pooled_projections token_replace_emb = None if self.image_condition_type == "token_replace": @@ -324,8 +323,9 @@ class HunyuanVideoConditionEmbedding(nn.Module): if self.guidance_embedder is not None: guidance_proj = self.time_proj(guidance) guidance_emb = self.guidance_embedder(guidance_proj.to(dtype=pooled_projection.dtype)) - conditioning = conditioning + guidance_emb - + conditioning = timesteps_emb + guidance_emb + pooled_projections + else: + conditioning = timesteps_emb + pooled_projections return conditioning, token_replace_emb