CivArchive
    Stable Video Infinity SVI (Wan 2.2 14B) - v2.0 Rank 64 Low Noise
    Preview 114204144

    Stable Video Infinity SVI (Wan 2.2 14B) — LoRA for Video Motion Enhancement

    This LoRA is a Wan 2.2–compatible adaptation of the motion patterns from Stable Video Infinity (SVI).
    It is designed for Image-to-Video (I2V) workflows in ComfyUI, especially on cloud environments that require LoRAs hosted on Civitai.

    The LoRA enhances:

    • Smooth temporal motion

    • Stability between frames

    • Coherent subject movement

    • Natural motion interpolation

    • Reduced jitter in I2V transformations

    This LoRA pairs extremely well with Wan 2.2 (14B) and works in both local ComfyUI and ComfyUI Cloud.


    How to Use (ComfyUI / Wan 2.2)

    • Load with a LoRA strength between 0.6 – 1.0

    • Works with both Power LoRA Loader and Load LoRA Model Only

    • Best results when combined with:

      • Wan 2.2 I2V base

      • Noise reference frames

      • CFG in the 1.0 – 1.5 range

    This LoRA does not alter subject appearance — it improves motion quality only.


    Source Project Attribution (Required by MIT License)

    This LoRA is derived from the Stable Video Infinity (SVI) research project by VITA-EPFL.

    Original Project:
    https://github.com/vita-epfl/Stable-Video-Infinity

    Original License: MIT License
    Stable Video Infinity and its pretrained models are distributed under MIT terms, which permit redistribution and modification with attribution.


    Additional Credits

    ComfyUI adaptation & workflow inspiration:
    Kijai

    Description

    The Rank-64 versions originate from the official SVI LoRAs hosted by Kijai (https://huggingface.co/Kijai/WanVideo_comfy/tree/main/LoRAs/Stable-Video-Infinity/v2.0).

    LORA
    Wan Video 2.2 I2V-A14B

    Details

    Downloads
    29
    Platform
    CivitAI
    Platform Status
    Deleted
    Created
    12/17/2025
    Updated
    12/19/2025
    Deleted
    12/19/2025

    Files

    SVI_Wan2.2-I2V-A14B_lora_LOW_v2.0_rank_64_fp16.safetensors