CivArchive
    Wan2.2 FLF2V (first last frame to video) for 3 (three) images - v1.0
    Preview 97017078

    Maybe not relevant thing, I barely search something like that.

    Well, it's just FLF2V workflow, based on lightx2v (6 steps) and 14B models, which can take three images and consistently turn them into short clip.

    Works good with a banch of similar looking pictures.

    Thinking about to do more pics than 3, but... Tried to merge clips together with avidemux and got nice result (1->2->3, 3->4->5 and more, same seed, for loop - 3->4->1 or 3->2->1 depends on pics). Got 40 seconds clip with visible, but not critic transitions. But, at last step I got kinda not very good part, so remade it with different seed, and than I got it: if I got 12 pics in one workflow - I'd just wasted time thanks to bad last clip. So, I think, better to do 3 pic clips and merge them elsewhere.

    Description

    FAQ

    Comments (4)

    maxtusrordey512Aug 28, 2025
    CivitAI

    can this work with loras?

    forfreelsd368
    Author
    Aug 29, 2025

    There is power lora loader in top left side, where you can add some loras under lightx2v lora, but on whole clip, not separately. They don't always work as expected, in the case of three pictures. For example, there are three pictures of a duck, you use lava lora to replace water. As a result, you get a mixture of water and lava in step 1->2, then just water again (because there was no lava in the 2nd picture), then again the process of mixing lava with water in step 2->3. I haven't experimented much, maybe everything is solved somehow, or seed jerking will helps, lora weights. One time I got result, where from 2nd picture Wan took only motion but not character, and it was like 1->3 transition (so, in that case, lava lora may work without interruption by 2nd picture).

    Besides, now there is one pack of lora for each step, maybe someday I will try to make separate ones, as well as a two positive prompt (or someone else will do it by then). Also, there must be a simplier way to do more steps, more pics (4, 5 ... ) and etc, than just cloning and repeating again and again whole second part of workflow.

    forfreelsd368
    Author
    Aug 29, 2025

    Or, you can make a new 2nd picture by taking shot from 1->2 clip where lora working like intendent and then 1->new2->3. Maybe, new 3rd also will ned, but from what I saw - usually very last frame not that much ruins everything.

    ranklens168Aug 29, 2025· 1 reaction
    CivitAI

    Thank you!

    Workflows
    Wan Video 2.2 I2V-A14B

    Details

    Downloads
    195
    Platform
    CivitAI
    Platform Status
    Available
    Created
    8/28/2025
    Updated
    5/15/2026
    Deleted
    -

    Files

    wan22FLF2VFirstLastFrameTo_v10.zip

    Mirrors