Choose text or image
Start from text for a fresh composition or from an image when the first frame already works.
VIDEO / CINEMATIC / 18+LTX
LTX
2.3
Create short fictional adult video scenes from a written direction or animate an existing generated frame. Choose the route, describe one cinematic beat and iterate with controlled changes.
OUTPUTAI VIDEO
MODESTEXT / IMAGE
FORMAT720P / SHORT CLIPS
FOCUSIDENTITY IN MOTION
LTX 2.3 / Video modelDirect a short scene from text or a source frame.
Choose one visual beat, one movement and one camera idea.
/ MODEL 04DIRECT THE
DIRECT THE
MOMENT.
LTX 2.3 supports two creative routes: text-to-video for building a scene from a written concept, and image-to-video when an existing fictional adult frame should anchor the composition and identity.
Short clips benefit from one visual beat. A clear motion verb, a simple camera instruction and a specific atmosphere give the model a readable sequence to solve. Iterate one variable at a time instead of rewriting every part of the direction.
Write the visual beat
Describe the subject, action, environment, pace and camera in a clear sequence.
Generate the short clip
Review identity, anatomy, direction of motion and continuity across the full result.
Iterate one variable
Adjust only the action, camera, timing or source frame so each change is measurable.
/ WHY LTXTWO WAYS
TWO WAYS
INTO THE SCENE.
Use text when the composition is still open; use an image when the visual identity is already decided.
This choice changes how the prompt works. Text-to-video carries more responsibility for the subject and environment. Image-to-video can spend more of the instruction on movement, timing and the camera because the first frame already supplies the visual foundation.
Text-to-video
Develop a new short scene directly from a compact cinematic description.
Image-to-video
Animate an eligible generated frame while preserving its visual starting point.
Short-form direction
Build the clip around one action and one readable emotional beat.
Controlled iteration
Compare focused changes to motion, timing, framing or atmosphere.
/ VIDEO EXAMPLESONE IDEA.
ONE IDEA.
THREE ROUTES.
The same concept can be directed through subject motion, environmental motion or camera motion.
Give one type of movement priority. When every layer moves aggressively at the same time, the scene loses a stable visual anchor.
01 / TextBuild the full moment.
Define the fictional adult, location, action, atmosphere and camera from the prompt.
02 / ImageAnimate an approved frame.
Let the image carry identity while the prompt concentrates on performance.
03 / CameraDirect the point of view.
A slow push, tracking move or locked composition changes how the action feels.
/ PROMPT GUIDEWRITE A
WRITE A
SHORT SEQUENCE.
Write the video prompt in the order the viewer experiences it.
Introduce the subject and environment, state the main action, describe its pace and end with the camera direction. For image-to-video, omit visual details already established by the source unless they must remain unchanged.
Reusable text-to-video structure
Original fictional adult character [description] in [setting]. The character [one action] at [pace] while [secondary natural motion]. [Lighting and atmosphere]. Camera [single movement or locked shot]. Clearly adult, no resemblance to a real person.Choose the route
Decide whether a written concept or a generated image is the better starting point.
Describe one beat
Give the short clip a focused action that can be read from beginning to end.
Keep the camera simple
Use one move that supports the action instead of competing with it.
Review the whole clip
Check the start, middle and final frame for identity, anatomy and visual continuity.
/ TEXT OR IMAGECHOOSE YOUR
CHOOSE YOUR
STARTING POINT.
Start from text for discovery. Start from an image for stronger visual control.
If the character and composition are not settled, text-to-video can reveal new scene directions. If an existing image already has the right identity and framing, image-to-video lets you focus on how that exact moment changes.
Two creation routesUse the input that solves the problem you actually have.
Discovery needs space; continuity needs a stable visual anchor.
/ FAQBEFORE YOU
BEFORE YOU
DIRECT.
Quick answers about the LTX 2.3 workflow.
What can LTX 2.3 generate?
It is presented here as a short-form video model with text-to-video and image-to-video routes for original fictional adults.
Should I start from text or an image?
Use text when you want the model to develop the composition. Use an image when identity, pose and framing are already established.
How long should the prompt be?
Long enough to define the subject, one action, pace, atmosphere and camera—without stacking unrelated events.
How do I improve identity continuity?
Begin from a clear generated source image, avoid contradictory instructions and keep movement within what the starting pose can support.
Why use short clips?
A short clip lets you evaluate one beat, make focused changes and avoid asking the model to maintain too many scene transitions at once.
Can I use an identifiable person's image?
No. Use original fictional adults only and never create deceptive or non-consensual sexual deepfakes.
/ START CREATINGDIRECT ONE
DIRECT ONE
STRONG MOMENT.
Choose text or an image, define a focused cinematic beat and refine the result through controlled changes.
Try LTX 2.3 ↗