Just tried that LTX 2.3 ingredient LoRA on my 6G VRAM, generating video straight from a concept art.

Just set up a new ComfyUI workflow using LTX 2.3’s image-conditioned LoRA. You can just throw in a reference design image and it generates a video, no more manually keyframing every single frame like before. The old director-based workflow was frame-by-frame grind, tedious and slow.

Now the approach is: first draw a character or concept design, then make the whole thing move. Way more intuitive. And I’m running this on just 6GB VRAM, which honestly surprised me.

Super friendly for character designers—you can see the motion right after finishing one version of the design, saving tons of time aligning frames. The workflow author said they’ll release it for free soon, so I’m waiting for that.

No way, 6G can actually run IC LoRA? That seriously blew my mind.

Being able to just take a concept art and make it move directly is such a time-saver.

Finally, the days of frame-by-frame motion tracking are over.

Dropping the workflow, when’s it coming?

Can this actually keep the character consistent?

Front row watching.