Nodes/ComfyUI-NL_Nodes/NL WanVideo ImageToVideo Encode v2 (multi-ref)
ComfyUI Node

NL WanVideo ImageToVideo Encode v2 (multi-ref)

A ComfyUI node in WanVideoWrapper with 18 inputs and 1 output.

By NOLABEL-VFX·Created 8 months ago·Updated 2 months ago· 0
NL WanVideo ImageToVideo Encode v2 (multi-ref)
  • vae
  • clip_embeds
  • reference_images
  • control_embeds
  • temporal_mask
  • extra_latents
  • add_cond_latents
  • image_embeds
width832
height480
num_frames81
noise_aug_strength0.000
start_latent_strength1.000
end_latent_strength1.000
force_offloadtrue
reference_frame_indicesauto
fun_or_fl2v_modeltrue
dual_encode_lastfalse
tiled_vaefalse
CategoryWanVideoWrapper

Inputs (18)

NameTypeDefaultDescription
widthINT83264–8096Width of the image to encode
heightINT48064–8096Height of the image to encode
num_framesINT811–10000Number of frames to encode
noise_aug_strengthFLOAT0.0000–10Strength of noise augmentation, helpful for I2V where some noise can add motion and give sharper results
start_latent_strengthFLOAT1.0000–10Multiplier for the first latent that receives a reference frame
end_latent_strengthFLOAT1.0000–10Multiplier for the last latent that receives a reference frame
force_offloadBOOLEANtrue
vaeoptWANVAE
clip_embedsoptWANVIDIMAGE_CLIPEMBEDSClip vision encoded image
reference_imagesoptIMAGEOne or more reference frames (start / middle / end)
reference_frame_indicesoptSTRINGautoComma separated frame indices for each reference. -1 targets the last frame. Use 'auto' to evenly spread references.
control_embedsoptWANVIDIMAGE_EMBEDSControl signal for the Fun -model
fun_or_fl2v_modeloptBOOLEANtrueEnable when using official FLF2V or Fun model
dual_encode_lastoptBOOLEANfalseUse Wan 2.2 style two-pass VAE encoding that treats the last frame as an end frame
temporal_maskoptMASKOptional per-frame mask; combined with reference locations and resized automatically
extra_latentsoptLATENTExtra latents to add to the input front, used for Skyreels A2 reference images
tiled_vaeoptBOOLEANfalseUse tiled VAE encoding for reduced memory use
add_cond_latentsoptADD_COND_LATENTSAdditional cond latents WIP

Outputs (1)

NameTypeDescription
image_embedsWANVIDIMAGE_EMBEDS