Wan2.2 Start Frame – End Frame Workflow (ComfyUI)

Use this ComfyUI workflow to create Wan2.2 AI videos from a start frame and an end frame, with control over motion between them and the frame rate; its output is a video.

The workflow takes image frames and text, applies Wan diffusion models and a VAE, then decodes and combines the result into a video.

Provide a start frame and an end frame; the workflow connects them in the generated video.

You can add a character LoRA to animate a custom AI character.

The published setup places diffusion models in ComfyUI’s diffusion_models folder, LoRAs in loras, the VAE in vae, and the text encoder in TextEncoders.

If you are new to ComfyUI, the publisher recommends starting with Comfy Manager and the Photorealistic ComfyUI workflow guide.

Required node packs listed for this workflow: comfyui-easy-use, comfyui-kjnodes, comfyui-videohelpersuite, and rgthree-comfy.

Required diffusion model files: wan2.2_i2v_high_noise_14B_fp8_scaled.safetensors and wan2.2_i2v_low_noise_14B_fp8_scaled.safetensors.

Other listed files include lightx2v_T2V_14B_cfg_step_distill_v2_lora_rank32_bf16, FastWan_T2V_14B_480p_lora_rank_128_bf16.safetensors, wan_2.1_vae.safetensors, and umt5_xxl_fp8_e4m3fn_scaled.safetensors.

Choose the high-noise or low-noise diffusion model.

The publisher advises setting the frame rate to 24 fps or 30 fps.

Sugerencia · sin verificar

For V2, the published instructions say LoRA models are not needed.

Before setup, verify where to obtain FastWan_T2V_14B_480p_lora_rank_128_bf16.safetensors; no URL is listed for this file.

Sugerencia · sin verificar

Before setup, verify the text-encoder filename: the description names umt5_xxl_fp16.safetensors, while the listed file is umt5_xxl_fp8_e4m3fn_scaled.safetensors.

Sugerencia · sin verificar