Introducing Lightricks LTX 2 Retake on WaveSpeedAI
Try Lightricks LTX 2 Retake for FREEIntroducing LTX-2 Retake: Surgical Video Editing That Preserves What Works
The era of complete video regeneration is over. Today, we’re thrilled to announce the availability of LTX-2 Retake on WaveSpeedAI—a groundbreaking AI model that lets you “reshoot” specific portions of your video without touching anything else. Think of it as having a virtual director’s chair with the power to say “let’s try that scene again” without calling the crew back or starting from scratch.
Whether you need to fix a single line of dialogue, shift a scene from day to night, or adjust an actor’s emotional delivery, LTX-2 Retake makes targeted video refinement as simple as describing what you want to change.
What is LTX-2 Retake?
LTX-2 Retake is Lightricks’ next-generation video refinement model, built on their acclaimed LTX-2 foundation—the first complete open-source AI video foundation model released in October 2025. While traditional AI video tools require you to regenerate entire clips when something needs adjustment, Retake takes a fundamentally different approach.
Using natural language prompts, you specify exactly which segment needs modification—“from 3 seconds to 7 seconds, make it nighttime”—and the model surgically edits just that portion. The surrounding footage remains untouched, with lighting, motion continuity, composition, and environment preserved throughout the transition.
This isn’t just a technical achievement; it’s a paradigm shift in AI-assisted video production. As noted by Scenario’s documentation, Retake “transforms the process from static one-off renders into an iterative, production-grade workflow.”
Key Features
LTX-2 Retake offers capabilities that set it apart from conventional video generation:
- Segment-level precision: Target specific time ranges (2-16 seconds) within your video for modification while preserving everything else
- Three replacement modes:
- Replace Video: Change visuals while keeping original audio intact
- Replace Audio: Regenerate dialogue, voiceover, or ambience while preserving visuals
- Replace Audio and Video: Update both picture and sound simultaneously
- Natural language control: Describe changes conversationally—no complex editing software required
- Motion continuity preservation: The model uses surrounding frames as context, ensuring seamless transitions in and out of edited sections
- 1080p output: Professional-quality resolution suitable for commercial and creative projects
- Identity preservation: When replacing dialogue or adjusting performance, the speaker’s identity and general style remain consistent
Real-World Use Cases
LTX-2 Retake addresses challenges that have long plagued video producers across industries:
Advertising and Marketing
Create multiple versions of an ad from a single master edit. Need a daytime version for one market and a nighttime version for another? Retake handles it without reshooting. Marketing teams can iterate on performance, tone, and pacing based on A/B testing results—all from the same original footage.
Film and Content Production
Post-production corrections become dramatically simpler. Fix specific dialogue lines that didn’t land without disrupting the actor’s otherwise perfect performance. Adjust emotional delivery in key scenes. Tweak camera movement or framing to better serve the story.
Corporate and Training Videos
Update voiceovers or ambience for localization without regenerating visual content. Refresh outdated information in specific segments while maintaining consistency with existing approved footage.
Social Media and Short-Form Content
Rapidly iterate on content based on performance data. If a specific moment in your video isn’t resonating, adjust just that portion rather than creating an entirely new piece.
As noted by RunDiffusion’s analysis, “It is cheaper to use LTX-2 Retake to edit a VEO 3.1 Fast than try to get the same generation again”—highlighting the efficiency gains for creators working across multiple AI video platforms.
Transparent, Predictable Pricing
WaveSpeedAI offers LTX-2 Retake at a straightforward rate of $0.10 per output video second:
| Video Duration | Billed Amount | Total Price |
|---|---|---|
| Up to 2 seconds | 2 seconds minimum | $0.20 |
| 5 seconds | 5 seconds | $0.50 |
| 10 seconds | 10 seconds | $1.00 |
| 16+ seconds | 16 seconds (capped) | $1.60 |
This pricing structure means you know exactly what each edit will cost before you commit. For a 10-second segment adjustment, you’re looking at just $1.00—a fraction of what reshooting would cost in time and resources.
Getting Started on WaveSpeedAI
WaveSpeedAI makes accessing LTX-2 Retake straightforward through our REST API:
- Navigate to the model page: Visit https://wavespeed.ai/models/lightricks/ltx-2-retake
- Prepare your source video: Upload your original clip (up to 20 seconds, under 100MB)
- Craft your prompt: Describe what should change and, optionally, specify the time range within your prompt
- Select your mode: Choose whether to replace video, audio, or both
- Generate: The API returns your modified video with the targeted changes applied
For best results, include in your prompt both what should change and what should remain the same. For example: “From 00:05 to 00:10, change the lighting to golden hour sunset while keeping the same actor, outfit, and camera angle.”
Why WaveSpeedAI?
Running LTX-2 Retake through WaveSpeedAI provides distinct advantages:
- No cold starts: Your API calls begin processing immediately
- Best performance: Optimized infrastructure for fast inference
- Affordable pricing: Pay only for what you generate, with no hidden fees
- Production-ready API: Enterprise-grade reliability for integration into your workflows
Expanding Your Video Editing Toolkit
LTX-2 Retake joins WaveSpeedAI’s comprehensive suite of video transformation tools:
- Video Head Swap: Replace entire heads (face, hair, outline) using reference images
- Video Face Swap: Swap facial identities while preserving expressions and lighting
- Video Eraser: Remove unwanted objects, people, or watermarks with temporal-aware inpainting
- Video Upscaler Pro: Upscale to 720p, 1080p, 2K, or 4K with exceptional frame consistency
Together, these tools form a complete AI video editing pipeline—from generation through refinement to final polish.
Conclusion
LTX-2 Retake represents a fundamental shift in how we approach AI video editing. Rather than the all-or-nothing proposition of traditional generation models, it offers the surgical precision that professional workflows demand. Fix what needs fixing. Keep what works.
The AI video analytics market is projected to reach $133.34 billion by 2030, and tools like LTX-2 Retake are driving this growth by making professional-grade video editing accessible, affordable, and fast.
Ready to transform your video editing workflow? Visit https://wavespeed.ai/models/lightricks/ltx-2-retake to get started with LTX-2 Retake today. With WaveSpeedAI’s no-cold-start infrastructure and straightforward pricing, you can begin reshaping your video content in minutes—not hours.
