LTX 2.3 IC-LoRA turns green-screen motion into a fully generated shot
waterarttrkgl · reddit · 2026-07-26
LTX 2.3 IC-LoRA regenerates a full shot from green-screen motion
A Reddit demo shows LTX 2.3 + IC-LoRA reconstructing a complete scene in ComfyUI from a green-screen source video used only as a motion reference.
- Pose sequence is extracted from the source footage and used as the control signal.
- A single first frame conditions the character, costume, environment, and lighting.
- The output is fully generated; there is no keying or compositing in the pipeline.
- According to the author, hand gestures and body timing transfer accurately, while only the motion timing comes from the source.
The post links to the workflow JSON in the Lightricks/ComfyUI-LTXVideo repository.
Related event: LTX 2.3 and IC-LoRA Enable Video Motion Reconstruction in ComfyUI(2 posts)→
More from Multimodal
- A neon-noir kaleidoscope prompt shows how to generate vivid fragmented AI art — azed_ai · 2026-07-26
- Quackchestra turns prompts into MIDI files with a chat interface and playable output — BLCNYY · 2026-07-26
- Opus 5 reportedly one-shots a playable FPS zombies game in a $400 prompt test — mattshumer_ · 2026-07-26
- Vivix A1 accepts voice, text and images mid-interaction, demo says — Scobleizer · 2026-07-26
- Vivix A1 demo shows synchronized speech, gaze, facial motion and body movement — Scobleizer · 2026-07-26
- A Reddit creator is two months into making mini AI films — sgtbb4 · 2026-07-26