Simple fast workflow for TikTok type of striptease videos good combo to use for images generated with Flux.Uses FastWan and lightx2v loras on 14B 480p fp8 WAN model which makes it very fast, takes 3-4 min to generate 4 sec video (480x832 97 frames) on 4080 16 GB VRAMDownload and use the loras otherwise the generated video will be blurry!!!- models/loras:<a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/FastWan/FastWan_T2V_14B_480p_lora_rank_128_bf16.safetensors">https://huggingface.co/Kijai/WanVideo_comfy/blob/main/FastWan/FastWan_T2V_14B_480p_lora_rank_128_bf16.safetensors</a><a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank32_bf16.safetensors">https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Lightx2v/lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank32_bf16.safetensors</a>I tested different versions of the loras, but these ones give the best resultsAdded slo-mo version with x2 interpolation which makes 6 sec 30fps videos which only adds 20 sec to generation timeAdded Steps Count and Split Options for easy changing the steps, you can try changing them to split step = 4 and step count = 8 for improved video quality but double the generation time (split = 3, count = 6 seems like good middle ground, but you can experiment depending on the image)You can use separate workflow for upscale and interpolate on the good video outputs as SwinIR takes more time to upscale than Wan takes to generate it :)

WAN 2.2 i2v 4 step nudify

<ul><li>clean up test nodes</li><li>added control for steps count and split</li><li>added slo-mo interpolation version</li><li>added upscale and interpolate video workflow</li></ul>

Files with SHA256: dfd3424fd188f7f9091d2a286c82a521b28fd2f019c87da28c85e7d8938ccf26