LTX 2.3 Sync

LTX 2.3 Sync packages LTX-2.3 into a more opinionated character animation workflow. Instead of treating it like a general video model, you give it a character reference, a motion source, and optionally an audio track. The Space preprocesses the reference video for pose-guided conditioning, can pull audio from that clip automatically, and turns the setup into a faster test for portrait animation and talking-head style motion transfer. A good first example is a clean front-facing portrait plus a short speaking clip with obvious head and hand movement.

Start with the Hugging Face Space and keep the first run simple: one portrait image, one short reference video, and the default settings. On the first pass, watch four things: whether identity roughly holds, whether the body motion follows the source clip, whether lip sync feels believable enough for a rough draft, and whether the result is useful before you touch prompt tuning. If it looks promising, the Space code is public and points to the fused checkpoint plus the upstream LTX-2.3 stack for a local path.

Treat this as a workflow pick, not a brand-new foundation model release. It is built on top of LTX-2.3, and the real test is whether identity, motion control, and lip sync stay usable on your own inputs rather than only on the Space examples.

• Animate a still portrait for rough avatar or character tests.
• Transfer body motion from a short clip onto a separate character image.
• Use a reference video's audio track to test quick talking-head style outputs.
• Pressure-test whether open motion-guided lip sync is good enough for social content or internal demos.

Try the demo

View model page