WEBVTT

00:00.207 --> 00:03.587
This is ComfyUI for Animation,
part five: In-betweening

00:05.955 --> 00:08.945
This workflow takes a sequence
of key frames and generates the

00:08.945 --> 00:12.455
in-between frames, producing
a complete character animation

00:14.635 --> 00:17.445
Load each keyframe into
the numbered image inputs.

00:17.835 --> 00:21.485
The default is five keyframes, but
you can bypass or duplicate the node

00:21.495 --> 00:23.425
groups to match your sequence length.

00:25.105 --> 00:26.945
Set your resolution width and height

00:29.379 --> 00:32.009
Then adjust the frame length
for each key frame pair.

00:32.499 --> 00:35.259
The recommended minimum is
seventeen frames per pair.

00:35.569 --> 00:38.519
Anything lower and the motion
becomes too short for the model

00:38.519 --> 00:40.119
to generate legible movement.

00:40.689 --> 00:44.279
The Select Every Nth Frame node
sits between each generation

00:44.289 --> 00:45.559
and the final assembly.

00:46.039 --> 00:49.089
If the motion between two key
frames is shorter than seventeen

00:49.089 --> 00:52.909
frames, increase this value to thin
out more frames from that segment.

00:53.429 --> 00:56.269
This is how you control the timing
in addition to the frame length.

00:58.169 --> 01:02.149
The workflow runs each consecutive
key frame pair through Wan 2.2's

01:02.149 --> 01:03.739
first and last frame model.

01:04.549 --> 01:08.039
The results from all pairs are then
thinned and assembled in sequence

01:08.279 --> 01:10.109
and rendered out as a single video