Hooked LTX-Video into ComfyUI to run video, this update is worth trying

The Forge / 工坊 Workflows / ComfyUI / 工作流
1

Jotting down my recent process of hooking LTX-Video into ComfyUI.

Wiring the nodes: this update’s ComfyUI support is pretty solid, the nodes connected up and ran with barely any hassle. If you’re already using comfy workflows, migration cost is basically zero.

Running drafts: it’s built for efficiency, the same job comes out noticeably faster than my old setup, perfect for drafts and quickly testing shots.

Pushing the parameters up: crank up the resolution and duration and VRAM demands kick in fast, don’t expect a finished piece in one shot.

So my workflow is settled now — use it first to lock down the shot composition and sense of motion, and once I’m happy, move on to refining it. Video generation has always been the direction I’m most obsessed with, and if you want to get into AI video, this path isn’t a bad entry point.

1 Like
3

I usually keep resolution really low at the draft stage, just to check if the motion is right first

4

Watching for VRAM usage, my card isn’t great

5

Thanks, could you go into more detail on how you wired the nodes?

6

I’m still stuck on the image-to-video path, haven’t run text-to-video much at all

7

Low entry barrier, I’m in :pray:

8

Lock the shots first then refine — do longer clips connect smoothly between segments?

9

I usually leave two or three overlapping frames at the join, otherwise the transition jumps a bit