WEBVTT

00:00.099 --> 00:04.419
This is ComfyUI for Animation,
part four: Concept Exploration.

00:05.999 --> 00:10.319
This workflow takes a concept art
illustration, generates a 3D proxy of

00:10.319 --> 00:14.939
the subject, and lets you rotate and pose
it freely before generating a polished

00:14.939 --> 00:17.179
still frame or video from any angle.

00:17.969 --> 00:21.679
Start by loading your concept art
of a prop into the Load Image node.

00:22.099 --> 00:25.959
Tripo generates a 3D model from
it and saves it as a GLB file

00:34.285 --> 00:38.865
Once that's done, bypass the previous
step and drag the 3D model into the Load

00:38.865 --> 00:43.765
3D node and set the material mode to
normal in the viewport's top left panel

00:50.531 --> 00:53.221
From here, the workflow
splits into two methods.

00:53.541 --> 00:57.991
For still frame exploration, pan and
rotate the viewport to the angle you want.

00:58.461 --> 01:02.481
The load 3D node outputs the current
view as an image automatically.

01:03.111 --> 01:06.391
That image, combined with your
original concept art and an optional

01:06.401 --> 01:11.391
background, goes into GPT Image
two to generate a polished on-style

01:11.391 --> 01:13.201
illustration at the new angle.

01:14.551 --> 01:18.301
For camera movement exploration,
click Start Recording in the load

01:18.351 --> 01:22.481
3D node, move through your camera
path, then click Stop Recording.

01:22.951 --> 01:26.581
The recorded video becomes the motion
reference for Seedance 2.0,

01:26.841 --> 01:30.841
which generates a fluid camera move
through the stylized scene using the

01:30.841 --> 01:33.051
subject and background as references.

01:34.711 --> 01:37.441
Bypass the method you're not
using, and the workflow runs

01:37.441 --> 01:39.021
cleanly through the one you need