CivArchive
    time lapse / Wan2.1 t2v 14B LoRA - beta
    Preview 65373708
    Preview 65373723
    Preview 65373729
    Preview 65373730

    AI time lapse of AI images drawn by AI.

    The fix version can be used with strength 1.0. Please set LoRA strength to 1.5 for the beta version.

    Please note, this model has a large capacity.

    It is designed for T2V with Wan2.1 14B model. It will not work properly with other models.

    This LoRA was trained on time lapse of three types of pictures I drew myself.

    https://x.com/8co28/status/1903652929481126020

    ::setting sample::

    512x768 (min 342x512~)

    ::prompt::

    time lapse,anime style,anime girl illust,1girl,drawing,2d,illustration,white background,static camera,brushstroke,

    ::negative::

    worst quality, low quality, JPEG residual, obscene, residual,hand,monochrome,character moving,photoreal,photo,realistic,

    Description

    The current version is a beta version. Please use it with LoRA strength set to 1.5.

    FAQ

    Comments (16)

    dennyekontoenMar 23, 2025· 11 reactions
    CivitAI

    If you made a dataset with time lapse videos that are going backwards and then trained a model on that, would that create a model that can remove detail from images? I think that would be interesting.

    hakoniwa
    Author
    Mar 23, 2025· 2 reactions

    That's interesting, worth a try.

    woctordhoMar 23, 2025

    Yes that's what Paints-UNDO originally proposed. I guess it's not too hard to train the lora for the I2V model

    xtoMar 24, 2025

    You could instead try setting an end frame using Kijai's nodes. Probably doesn't work as well as your suggestion though.

    ShurikMar 24, 2025

    Is that different than just reversing the video?

    dennyekontoenMar 24, 2025· 2 reactions

    @Shurik  I want to use it as I2V on art that wasn't made with this LORA.

    ShurikMar 24, 2025

    @dennyekontoen Oh, I see. That makes sense.

    jkntMar 24, 2025

    @dennyekontoen and then reverse that to make a time lapse? actual 5head move lmao

    Hugs288Mar 24, 2025· 3 reactions
    CivitAI

    this is cool but I2V would be way more useful imo

    TempMar 24, 2025· 4 reactions
    CivitAI

    1.14gb, ouch, got any smaller versions?

    woctordhoAug 15, 2025

    Rank 128 is definitely too large. This can be pruned to 25 MB (mostly rank 2 or 3) while keeping 98% Frobenius norm, see https://huggingface.co/woctordho/wan-lora-pruned/blob/main/852wa_time_lapse_fix_wan2.1t2v_r128_fro0.9.safetensors

    HenkeiMay 13, 2025
    CivitAI

    grifters' favorite

    a1161327317Jul 27, 2025
    CivitAI

    Does anyone have a LoRA that can enhance the anime effect? How can I avoid the real - life effect in i2v?

    woctordhoAug 15, 2025· 2 reactions
    CivitAI

    This can work with the FLF2V (First and Last Frame to Video) workflow. Put a completely white image as the first frame, your desired image as the last frame, then it will generate the time lapse.

    RsmbGrantAug 17, 2025

    wonderful

    RsmbGrantAug 17, 2025
    CivitAI

    OP can you please make one for ltxv? i have a good low vram workflow i use and fast for my needs. But I cant use Wan oviously sadly. ( Ltxv 2b 0.9.8) Thanks <3

    LORA
    Wan Video

    Details

    Downloads
    135
    Platform
    CivitAI
    Platform Status
    Available
    Created
    3/23/2025
    Updated
    5/22/2026
    Deleted
    -
    Trigger Words:
    time lapse,anime style,drawing,2d,illustration,girl,static camera,brushstrokes,coloring,

    Available On (2 platforms)

    Same model published on other platforms. May have additional downloads or version variants.