- Clean Plate LoRA for LTX-2.3 removes people, vehicles, and cast shadows from footage, then inpaints the background in one pass with no manual rotoscoping.
- Trained on real-world footage across varied weather, angles, and lenses, so it handles shadows and unexpected scenarios with consistent lighting and texture.
- Controlled by a positive prompt describing the empty location, detailed negatives listing what to exclude (people, shadows, ghosting), and adjustable LoRA strength.
Removing people from video footage is one of the hardest post-production problems. Shadows cast by actors, moving vehicles, and the subjects themselves leave traces that simple masking can't handle. The Clean Plate LoRA for LTX-2.3 solves this: remove people, vehicles, and their shadows, then intelligently inpaint a seamless background in a single pass.
See It In Action
Example 1: Before/After Comparison
Example 2: Input and output
Input:
Output:
What It Does
A clean plate is a background plate free of moving objects, actors, and shadows. In traditional VFX, creating one requires either reshooting without talent or painstaking manual rotoscoping and inpainting.
The Clean Plate LoRA removes people, vehicles, and cast shadows from video footage, then fills in the background. The result is a usable clean plate ready for compositing or as a standalone background element. No manual masking. No per-frame rotoscoping. One prompt, one video, one clean background.
Intelligent inpainting alone often produces artifacts or inconsistent lighting when shadows are involved. This LoRA was trained specifically on real-world footage with people, vehicles, and moving shadows across varying weather, camera angles, and lenses. That diversity means it handles unexpected scenarios reliably.
The result is seamless background recovery with consistent lighting and texture. Perfect for visual effects, environment documentation, and background cleanup.
How It Works
The LoRA uses inpainting conditioning to understand the scene geometry and lighting, then removes specified objects while preserving the background coherence.
This requires a clear description of what you want removed and what should stay. The positive prompt describes the target state: an empty version of the exact same location, with all its environmental details intact. The negative prompt lists what should not appear: people, vehicles, body parts, shadows, and any artifacts that indicate incomplete removal.
Here's a concrete example:
Positive: An empty clean plate of the exact same location: a modern open-plan office with identical desks, office chairs, monitor dividers, concrete pillar, exposed ceiling with pipes and acoustic panels, large windows with city view, floor and lighting as the source video. No people, no humans, no figures, and no body parts such as faces, arms, hands or legs anywhere in the frame. No backpack, no ID badge, no lanyard anywhere in the frame. No cast shadows from people on the floor or walls. Static photorealistic footage, natural indoor light, high detail.
Negative: person, people, human, man, figure, face, beard, glasses, arm, hand, leg, body part, jacket, hoodie, shirt, backpack, bag, id badge, lanyard, name tag, shadow of a person, human shadow, cast shadow, leftover person, leftover figure, ghosting, semi-transparent remnants, blurry, soft, distorted, inconsistent motion, jitter, worst quality, cartoon, video game, ugly
LoRA strength: 1.00
The detailed negatives matter. By explicitly listing what shouldn't appear (shadows, ghosting, semi-transparent remnants), you're teaching the model what clean removal looks like.
The Office Worker example above uses a LoRA strength of 1.00. Increase the strength if removal is incomplete; decrease if the inpainting becomes too aggressive or unnatural.
When to Use It
- VFX cleanup for compositing reference plates
- Environment documentation without moving people
- Background recovery for green-screen compositing
- Location footage preparation for post-production
- Footage repurposing where background environment matters more than talent
Any scenario where you need the background without moving subjects or their shadows.
Get Started
The LoRA is available on Hugging Face. Join the community in Discord and Instagram for announcements on new releases and shared prompts.
Load your source video. Describe the clean scene and what to remove. Let LTX-2.3 handle the inpainting.
One video. One prompt. One clean plate.
