Why creators choose LTX-2.3

LTX-2.3 is the 4K workhorse from Lightricks. It reaches full 4K natively, includes sound, extends video at 1080p by up to 20 seconds, and keeps characters consistent across scenes. It's built for multishot production where character consistency across cuts matters. When you need broadcast-resolution video with multishot character continuity, LTX-2.3 delivers.

Native 4K without upscaling

LTX-2.3 understands 4K from the ground up. No AI upscale, no quality loss. You get the resolution the model trained on. For broadcast and commercial work, native 4K is the spec. LTX-2.3 hits it in a single pass from text or image input.

Synchronized audio every time

Sound is built into the generation. No timing problems, no separate audio pipeline. The dialogue and effects sync to the motion because they're made together. For projects where audio is part of the creative, this is essential. No async issues to fix in post.

1080p video extension up to 20 seconds

Most models cap extension at 10 seconds. LTX-2.3 extends up to 20, so longer continuations stay in one pass. You can extend a shot with significant new action without cutting to a new generation. That means fewer seams and more flexibility in the edit.

Multishot character generation

When you're building a scene, LTX-2.3 keeps the same character across multiple shots. Use it to generate an establishing wide, then a close-up on the same person, and the faces will match. This is critical for narrative work and character-driven content. The model is trained for this specific use case.

How to use LTX-2.3 in Craze

1. Start a project and describe the scene or add reference images. If you have video to extend, upload it.
2. Say you want LTX-2.3. Craze runs the model when you ask for it by name. If you don't name one, Craze shows compatible models with their costs.
3. Craze generates the clip at your chosen resolution with sound included.
4. Review in the editor. Regenerate shots as needed or continue building your scene.

For multishot work, describe the full scene and Craze breaks it into shots, using LTX-2.3's character consistency to keep the same person across cuts.

LTX-2.3 capabilities

LTX-2.3 supports text-to-video, image-to-video, and video extension. You describe a scene and it generates from scratch. You provide an image and it animates it. You provide video and it extends it with optional audio that continues the story.

The model is built for narrative work and character-driven content. It keeps characters consistent within a scene and across multiple shots. It handles complex scenes with multiple subjects, precise motion, and realistic physics. LTX-2.3 generates at 1080p or 4K, with sound always included in the output.

Lightricks released LTX-2.3 under an Apache 2.0 open license, meaning the model weights are public. On Craze, you just use it by name and Craze handles the integration. You don't need to think about deployment or licensing. Craze runs it on your behalf.

LTX-2.3 pricing

You pay per second of video. Extension starts at $0.106 per second of added video and is available at 1080p only; new text-to-video and image-to-video clips have separate resolution-dependent rates. Sound is included in every generation.

The cost calculator on this page shows exact rates for any combination of resolution and clip length. The 20-second extension capability means you can often accomplish longer narratives without generating multiple clips to stitch together.

Who LTX-2.3 is for

Broadcast and commercial work

Native 4K is the standard for commercials and broadcast. LTX-2.3 hits that spec in one pass. For production studios and agencies, this means no post-upscale and no quality loss to that specification.

Narrative sequences with consistent characters

Build a scene where the same character appears in wide, medium, and close-up shots. LTX-2.3 keeps the person recognizable across all three. For storytelling, this consistency is critical. Audiences can follow a character when they're consistent across cuts.

Projects where sound is part of the creative

Dialogue, ambience, and effects sync perfectly because they're generated together. No audio timing work. For narrative work and character-driven content, having audio baked in means one unified creative decision instead of two separate problems.

Multishot narrative production

When you're building a scene with multiple camera angles of the same character, LTX-2.3 is built for this workflow. Generate establishing shots, then tighter framing, then close-ups, all with the same character. The model maintains consistency across all of them.

LTX-2.3 vs other models

LTX-2.3 is the 4K choice with strong character consistency and multishot support. For 4K speed optimization, see Veo 3.1 Fast, which reaches 4K but generates faster. For character consistency with flexible reference images, see Gemini Omni Flash. For longest single clips at 20 seconds but lower resolution, see FLUX.3. For talking photos and audio-to-video, see LTX-2.5, the newer version with additional capabilities.