CivArchive
    Wan2.2 i2v gguf + lightx2v lora - i2v_v1
    NSFW
    Preview 94548890

    My simple 4steps workflow which I use to generate videos from image on my rx 6800xt graphics card.

    To increase speed Lightx2v Lora must be installed. Generation of 33 frames video takes about 250 seconds after the models are fully loaded.

    In my opinion lightx2v for wan2.1 provides much better results than version for wan2.2 so I use first one.

    All required models download:

    Wan2.2-I2V-A14B-HighNoise-Q3_K_S.gguf

    Wan2.2-I2V-A14B-LowNoise-Q3_K_S.gguf

    umt5_xxl_fp8_e4m3fn_scaled

    wan_2.1_vae

    Wan21_T2V_14B_lightx2v_cfg_step_distill_lora_rank32

    Required custom nodes:

    GGUF

    Rgthree

    VideoHelperSuite

    Description

    FAQ

    Comments (5)

    ZealotAug 17, 2025
    CivitAI

    KSamplerAdvanced

    Given groups=1, weight of size [5120, 36, 1, 2, 2], expected input[1, 32, 9, 80, 60] to have 36 channels, but got 32 channels instead

    ternieoffAug 18, 2025

    i always get this to all gguf workflows. idk how to fix it

    ZealotAug 18, 2025· 1 reaction

    ternieoff solved, update the comfyui verison

    gaming_01Dec 8, 2025

    @Zealot ya lo actualize y aun tengo el mismo problema

    christianmontagx127Sep 14, 2025· 1 reaction
    CivitAI

    Works great. Thank you! I just replaced the tiled Vae decoder to the normal one, because the tiled one does a lot of flickering in the final video.

    Workflows
    Wan Video 14B i2v 480p

    Details

    Downloads
    1,244
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/16/2025
    Updated
    4/30/2026
    Deleted
    -

    Files

    wan22I2vGgufLightx2v_i2vV1.zip

    Mirrors