LTX-2 Retake: Directable AI Video Editing for Creators on RunDiffusion

LTX-2 Retake lets creators regenerate selected moments within a video while preserving continuity. You can reshape tone, dialogue, or motion without re-rendering entire scenes, making Runnit a powerful tool for directable AI filmmaking.
LTX-2 Retake: Directable AI Video Editing for Creators on RunDiffusion

LTX-2 Retake is a new AI-driven editing feature designed to give creators, directors, and production teams unprecedented control over how a moment unfolds within a video. Instead of fixing small issues frame-by-frame, Retake allows you to reshape intent, pacing, movement, emotion, or dialogue while keeping the rest of the shot perfectly intact. In this guide, we’ll cover how Retake works, why it matters for modern video workflows, and what you can expect when it launches on Runnit.


What Retake Does

Retake regenerates a selected time range within a shot while preserving continuity with the original footage. You can adjust:

  • Beginning or ending performance
  • Camera motion or framing changes
  • Acting choices, gestures, or emotional tone
  • Dialogue phrasing while maintaining the same voice
  • Narrative direction or pacing

Instead of starting from scratch, Retake builds on the footage you already have allowing creators to refine the exact moment that needs improvement, while keeping the rest of the scene consistent.

It is particularly useful in workflows built around AI videos with RunDiffusion, where rapid iteration and feedback cycles are essential.


Who Retake Is For

Retake is built primarily for creative professionals:

  • Directors
  • Agencies
  • Post-production teams
  • Motion designers
  • Content creators
  • Animation and pre-viz departments

But the feature also unlocks creative flexibility for any LTX-2 user. Whether you’re correcting tone, shifting story direction, refining continuity, or improving performance, Retake makes it possible to change a moment without throwing away a great shot.


How Retake Works (High-Level)

Retake uses a targeted AI regeneration system:

  1. You select a time range within the input video.
  2. LTX-2 regenerates that segment entirely from noise, ensuring new motion, acting, or dialogue is consistent with your prompt.
  3. The model strongly attends to surrounding frames, blending the regenerated moment seamlessly into the original footage.
  4. The rest of the video remains untouched.

This is similar to the “Shot Fix” concept in video inpainting but optimized for continuity, motion coherence, and directorial control.

For users who already work with advanced pipelines like the LTX Video Workflow for ComfyUI, Retake brings this editing precision into a streamlined, cloud-based workflow.


Capabilities & Creative Applications

Retake isn’t just a correction tool it’s a directorial feature. You can:

  • Reimagine an emotional beat
  • Adjust an actor’s expression or gesture
  • Rephrase dialogue while keeping the same character voice
  • Modify tone to match narrative changes
  • Improve timing or pacing
  • Correct or stylize motion
  • Shift the story direction without re-shooting

Early use-cases include:

  • Brand videos needing tone corrections
  • Social content requiring quick storyline adjustments
  • Pre-visualization where shots evolve rapidly
  • Cinematic sequences needing refined emotion or timing
  • Architectural or product videos needing updated context or framing
    (pairing well with turning ideas into architectural videos)

Retake keeps continuity intact while giving you control over the moment itself—the first of many directable features coming to LTX-2.


Technical Limits & Requirements

Upload Limits

  • File size: under 100 MB
  • Max input duration: 20 seconds

Edit Segmentation

  • Up to 505 frames (≈ 20 seconds @ 25 fps)
  • Max edit range: 16 seconds
  • Min edit range: 2 seconds (moving to 2.5 seconds soon)

Output

  • 1080p output only (4K uploads are downscaled)
  • Additional resolutions coming soon

Workflow Overview on RunDiffusion

Using Retake on Runnit follows the same streamlined pattern as other video tools:

Go to the Runnit Homepage

On the left sidebar Select Generate then Video

Choose LTX-2 Retake from the tool list

Upload your footage

Select your editable time range

Provide direction, such as:

    • Emotional tone
    • Performance notes
    • Dialogue phrasing adjustments
    • Camera motion changes

Generate and review your updated video

Iterate immediately if needed

If you work with multiple Runnit tools (e.g. editing, animating, reshooting), Retake fits smoothly into modern workflows, especially alongside video tools such as LTX-2 and our suite of Video models.


Why Retake Matters

Retake introduces something new to AI filmmaking:
the ability to direct, not just generate.

Traditional AI video systems required re-rendering entire shots to make small changes. Retake solves this by letting you:

  • Preserve what works
  • Modify only what you need
  • Act on feedback immediately
  • Iterate faster and with greater creative precision

It helps creators move from “close enough” to intentional, directable storytelling, where nuance, pacing, and emotion truly matter.

About the author
Adam Stewart

Great! You’ve successfully signed up.

Welcome back! You've successfully signed in.

You've successfully subscribed to RunDiffusion.

Success! Check your email for magic link to sign-in.

Success! Your billing info has been updated.

Your billing was not updated.