WAN 2.2 5B Turbo Checkpoint T2V/I2V. All included: U-Net, CLIP and VAE. Just load as any other checkpoint model and have fun: <img src="https://image.civitai.com/xG1nkqKTMzGDvpLrqFT7WA/63871b18-9a8e-4d1d-96ff-c26ae02fcb22/width=525/63871b18-9a8e-4d1d-96ff-c26ae02fcb22.jpeg" /><code>4</code> steps is enough. <code>CFG 1</code>, sampler is <code>Euler</code> or <code>SA_Solver</code> or <code>Uni_PC</code>, scheduler is <code>simple</code> or <code>normal</code> or <code>beta</code>. Multiplier <code>0.8</code> applied to latent will help with oversaturation. Preferrable sizes are 1280x704 and 704x1280.<a target="_blank" rel="ugc" href="https://civarchive.com/search/models?baseModel=Wan%20Video%202.2%20TI2V-5B&amp;modelType=LORA&amp;sortBy=models_v9">LoRAs compatible with this model on Civitai</a>.Sample workflow attached here as <a target="_blank" rel="ugc" href="https://civarchive.com/api/download/models/2258309?type=Training%20Data">train data</a>. Tech info:<ul><li>base model is <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy_fp8_scaled/blob/main/TI2V/Wan2_2-TI2V-5B_fp8_e4m3fn_scaled_KJ.safetensors">Wan2_2-TI2V-5B_fp8_e4m3fn_scaled_KJ</a> 1. fp8 Fast is merged with <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/FastWan/Wan2_2_5B_FastWanFullAttn_lora_rank_128_bf16.safetensors">Wan2_2_5B_FastWanFullAttn_lora_rank_128_bf16</a> 2. fp8 Turbo is merged with <a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan22-Turbo/Wan22_TI2V_5B_Turbo_lora_rank_64_fp16.safetensors">Wan22_TI2V_5B_Turbo_lora_rank_64_fp16</a> </li></ul>

Wan Damme - Rapid WAN 2.2 5B 4-steps Checkpoint T2V/I2V

<a target="_blank" rel="ugc" href="https://huggingface.co/Kijai/WanVideo_comfy/blob/main/Wan22-Turbo/Wan22_TI2V_5B_Turbo_lora_rank_64_fp16.safetensors">Wan22_TI2V_5B_Turbo_lora_rank_64_fp16</a>

Files with SHA256: bdf342bfcd6608f2252ea1864ecb527c1ad337b6a7cbfbe4b071447d886c6b4a