LTX 2.3 · Obscura Object Removal from Video

HOW IT WORKS

Step 1. Upload your video The clip you want to clean up. Make sure the thing you want gone is clearly visible in the shot. The cleaner the view of it, the cleaner the erase. Works great with: static shots · mild camera motion · single clear object

Step 2. Describe what to remove Write a short prompt that points at the object. "Remove the chair from the background" reads better than "remove the chair." Be specific so the model knows where to look.

Step 3. Hit run and download You get back an MP4 with the object erased and the rest of the scene held in place. Audio is optional, so toggle it on if you want the original sound. Ready for: Premiere · DaVinci Resolve · Final Cut · any editor

First time? Leave every setting as-is. The defaults (720 x 1280 · 81 frames · 24 fps · random seed) are the right starting point for almost everyone.

RECOMMENDED SETTINGS

Quick-start guide. Find the goal that matches yours and copy the settings.

• Standard cleanup (most people) Start here. Defaults across the board, LoRA strength 2.3, audio off. The right starting point for almost everyone.

• Result looks over-edited. Lower the LoRA strength below 2.3 in small steps. The erase comes through softer and the surrounding scene keeps more of its original texture.

• Keep the original sound. Turn on Use Original Audio. The source clip's audio carries through to the export.

• Reproduce a result you liked. Lock the seed to the number that produced it. Same seed and same prompt gives you the same cut again.

• Vertical or landscape footage. Set Resolution to match your source. 720 x 1280 for vertical, swap the values for landscape.

• Longer clip. Raise the Length above 81 frames. More frames means a longer result and a longer generation time.

Prompt: Describe the one element you want gone and where it sits in the frame. Keep it short and specific. Naming the location ("in the background", "on the left") helps the model find the right object.

Negative prompt: What to avoid. The default handles common artifacts, so leave it alone on a first run and revisit it only if something unwanted shows up.

USE CASES

🎬 Editors & Post Teams Pull a stray person, a passing car, or a piece of gear out of a shot without a manual paint-out pass. Describe it, run it, and check the result before committing to a full cleanup.

🎥 VFX Artists Skip the first rotoscope pass. Remove a boom mic that dipped into frame, a crew reflection, or a rig that crept into the edge of the plate, then take the clean version into your comp.

📱 Content Creators Clean up footage you already shot on your phone. Drop a distracting object out of the background of a talking-head clip or a product shot without reshooting.

🏢 Brand & Agency Teams Clear a logo, sign, or branded item you cannot license out of client footage. Point at it, run it, and deliver a clean cut.

WHAT WORKS BEST / WHAT TO AVOID

✅ Works great

• Object clearly separated from your subject

• Static shots or mild camera motion

• A single clear element to remove

• Well-lit footage

⚠️ May produce softer results

• Object touching or overlapping the subject

• Fast, complex camera movement

• Crowded scenes with many overlapping objects

• Extreme low-light clips

NEW TO COMFYUI?

Start with the free ComfyUI for Beginners Course on Floyo. Sixteen short videos take you from zero to running your own AI workflows. No setup headaches, no jargon, clear hands-on lessons. Watch the course, then run any workflow here in your browser.

👉 Watch the free ComfyUI for Beginners Course →

FAQ

What is the Obscura Removal LoRA in LTX 2.3? It is a LoRA trained to erase objects from video while preserving the rest of the scene. Paired with the LTX 2.3 distilled transformer from Lightricks, it reads what you describe in the prompt and removes that element across every frame, holding the lighting, motion, and background steady.

Why does my LTX 2.3 removal look over-edited? The LoRA strength is likely too high for your clip. The default of 2.3 pushes hard. Lower it in small steps and the erase softens, which lets the surrounding scene keep more of its original detail. Stop when the result reads natural.

Can I keep the original audio on the output video? Yes. Turn on the Use Original Audio switch and the source clip's sound carries through to the export. It is off by default, so leave it off if you only need the cleaned visuals.

How long a video can LTX 2.3 process at once? Length is set to 81 frames by default and you can raise it. More frames means a longer clip and a longer generation time. For longer footage, split it into segments and run the cleanup on each.

Floyo vs RunPod for running this LTX 2.3 workflow? RunPod rents you a GPU pod or a serverless endpoint, then you set up ComfyUI, install the custom nodes, place the model files, and often build a Docker image yourself. That gives you infrastructure control. Floyo skips all of it. This workflow is already configured, so you load a clip in your browser and run.

Can I run LTX 2.3 and the Obscura LoRA on Weavy or Figma Weave? Weavy, now Figma Weave, runs its own hosted models on a node canvas built for pairing generation with editing tools. It does not load community ComfyUI workflow files or arbitrary LoRAs. Floyo runs this exact ComfyUI graph, LTX 2.3 plus the Obscura Removal LoRA, with the real node settings exposed.

Floyo vs Flora for video object removal? Flora is a node-based canvas aimed at ideation and experimentation with a curated set of models, and it tends to hide granular controls. It does not run community ComfyUI graphs. Floyo runs the actual workflow, so you get the same LoRA strength, seed, and resolution controls the builder set up.

Does Civitai run this kind of LTX 2.3 video removal workflow? Civitai is mainly a model and LoRA repository with an on-site generator focused on image generation. It is not built to run a full ComfyUI video pipeline like this one. You would download assets there and run them elsewhere. On Floyo the assembled workflow runs directly in the browser.

How to run LTX 2.3 online? You can run LTX 2.3 online through Floyo. No installation, no setup, no GPU to rent. Open the workflow in your browser, load your video, write a prompt, and hit run. Free to try.

WHY FLOYO?

Floyo is the only platform with team collaboration for ComfyUI in the browser. You run workflows with no install. You share run history, assets, and models across your team. You pay only when you generate. Floyo supports open-source and closed-source models.

An editor runs a cleanup and likes the result. A teammate opens that exact run from shared history and keeps going. No file handoffs. No version confusion.

For studios and enterprise teams, Floyo adds private workspaces, pooled resources, and a team usage dashboard. Other ComfyUI cloud tools run for one person at a time. Floyo runs for the whole team, with transparent per-generation costs.

Ready to try it? Load your first clip and run it. The prompt and settings are already set.

→ Launch Workflow, Free

Questions? Watch the free course or check the FAQ above.