LTX 2.3 supports Depth-mode remakes from one base video

• egeberkina's main walkthrough frames LTX 2.3 as a video-to-video remake workflow: feed one base clip into LTX Studio, keep motion with Pose, Depth, or Edges, and rebuild the scene from a new prompt and start image.
• In egeberkina's setup clip, the base footage starts as a Nano Banana Pro still that gets turned into a 10-second reference video in Kling 3.0 before it ever reaches LTX.
• egeberkina's start-frame step shows the real lever: swap subject, character, or visual style in the opening frame while keeping the same composition skeleton.
• According to egeberkina's LTX-2.3 settings post, the actual remake only needs three inputs inside Video-to-Video: a new starting frame, the original clip as reference, and a control mode, in this case Depth.
• egeberkina's result clip and the final thread post both point to the same payoff: the shot keeps its timing and camera movement while the character design and scene identity get rebuilt.

You can watch the original thread opener, see the exact prompt and Depth settings, and jump straight to the LTX Studio workflow link. The odd, useful part is how little of the original clip needs to survive. egeberkina's alternate start frames keep the composition, then the output demo treats the rest like controllable style transfer for motion.

The workflow starts outside LTX. In egeberkina's setup clip, a Pilates image made in Nano Banana Pro becomes a 10-second motion reference inside Kling 3.0.

That reference clip does two jobs at once. It locks in pacing, camera movement, and pose continuity, and it gives the later LTX pass a stable motion template to rebuild around.

egeberkina's post reduces the remake step to one practical trick: generate multiple opening frames that change the character, subject, or style, but keep the same composition structure.

That turns the workflow into a reusable shot format. One motion source can support several visual directions, as long as the new first frame still matches the original scene geometry closely enough to guide the model.

In egeberkina's LTX-2.3 settings post, the actual Video-to-Video setup is just three actions:

1. Upload the new starting frame.
2. Add the original clip as the reference video.
3. Select Depth as the control mode.

The prompt matters because it tells LTX what to preserve and what to swap. egeberkina's example explicitly asks for the exact motion, timing, pacing, and camera movement from the reference clip while forbidding photorealism, style drift, morphing, and any conversion of the cartoon character into a realistic human.

egeberkina's result demo is the cleanest summary of what this is good at: keep the cinematic movement, then rebuild the shot's visual identity by changing the start frame plus control mode.

For creative work, that lands somewhere between animation retargeting and style transfer. The difference is that the structure comes from video controls, not just from a prompt, so the remake keeps more of the original blocking and camera logic.

The first post in the thread, egeberkina's opener, lists three preservation modes in LTX 2.3: Pose, Depth, and Edges.

Only Depth gets demonstrated here, but the menu hints at three different kinds of remakes:

• Pose for body-position fidelity.
• Depth for scene structure and spatial layout.
• Edges for contour-driven shot rebuilds.

The thread ends with egeberkina's final post linking directly to the workflow page in LTX Studio, which makes this less of a concept demo and more of a ready-made template people can start from.