Preview for Optimized WAN 2.1 Image-to-Video (I2V) 720P SageAttention + TeaCache

Run this workflow on InstaSD

Get started in minutes! Run this ComfyUI workflow online - no setup required.

Description

Optimized WAN 2.1 Image-to-Video (I2V) 720P workflow using SageAttention and TeaCache for significantly improved speed and efficiency.

šŸ”¬ Benchmark Results (RTX 4090 | 33 Frames | 16 FPS | 20 Steps)

  • ā±ļø Base: 542.09s
  • āš” Optimized (SageAttention + TeaCache): 248s
  • šŸ”„ 54% Speed Boost!

āœØ Key Features

āœ… Drastically Faster Inference ā€“ Cut generation time in half!

āœ… SageAttention Optimization ā€“ More efficient attention mechanisms

āœ… TeaCache Acceleration ā€“ Reduces redundant computations

Models

FileDestinationSource
wan_2.1_vae.safetensors/ComfyUI/models/vaeDownload
clip_vision_h.safetensors/ComfyUI/models/clip_visionDownload
umt5_xxl_fp8_e4m3fn_scaled.safetensors/ComfyUI/models/text_encodersDownload
wan2.1_i2v_720p_14B_bf16.safetensors/ComfyUI/models/diffusion_modelsDownload

Nodes

CLIPLoaderCLIPTextEncodeWanVideoTeaCacheKJPathchSageAttentionKJUNETLoaderKSamplerVAELoaderVAEDecodeTiledNoteCLIPVisionLoaderCLIPVisionEncodeLoadImageVHS_VideoCombineWanImageToVideo