>>29358180
Regarding this.
>>29358460 is two gens of i2v, one of them reversed, then a FLF, then a trip through VACE to clean up the loop.
>start with normal i2v workflow, gen 2x
>reverse one of them
>combine with a fast workflow that drops the last frame of the first and the first frame of the last, which in this case should be the same, and then interpolates the remainders to replace.
>judge that to be good enough in this case - if not, can run the two gens through a VACE workflow that automatically bridges a gap of a length I can specify
>FLF the resulting combo with itself
>splice that, same decision tree as before, didn't need the VACE bridge on this one either
>FLF didn't quite get back to the start - VACE workflow to smooth out the loop
>upscale/interpolate the result
So, 3 gens. Base image is at 5 second mark.
The VACE workflows were a bitch to set up, but they automatically create masks according to how long I specify the gaps to be, so it's easy. They're 1.3b, as the 14b VACE is too much for my humble machine, and for a 4~8 frame gap take about 3 minutes, vs the 30s the interpolate-splicer takes.
