LTX Video Prompt Guide: Write One Flowing Paragraph It Can Follow

You wrote an LTX Video prompt the way you’d write for most models — labeled sections, comma-separated tags, a list of attributes — and the output came back muddy and incoherent. That structure works elsewhere. With LTX Video it fights the model. LTX wants a single flowing paragraph that reads like a shot description, and switching to that one habit fixes most bad generations.

LTX Video is Lightricks’ open-source diffusion-transformer video model, notable for generating video in real time — 30fps output faster than playback speed. It runs locally on consumer hardware, comes in sizes from 2B up to 13B, and newer versions add synchronized audio and ControlNet support for depth, pose, and edge guidance. But the model’s defining quirk is how it reads prompts. This LTX Video prompt guide covers that quirk, what the model does best, and gives you example prompts written the way LTX expects.

What LTX Video Does Well

Knowing the model’s strengths tells you what kind of shots to prompt for:

• Real-time generation. LTX produces video faster than playback on capable hardware, which makes rapid iteration genuinely fast. You can test a prompt, adjust one detail, and regenerate without long waits.
• Grounded, everyday motion. LTX performs best with natural, realistic actions rather than fantastical or heavily stylized scenes. A person pouring coffee reads better than a dragon erupting from a volcano.
• Flexible output. It supports a range of aspect ratios — 16:9, 9:16, 1:1, 21:9 and more — and durations up to around 10 seconds, so it adapts to social formats and widescreen alike.
• Image and keyframe workflows. Beyond text-to-video, LTX handles image-to-video, keyframe-based animation, and video extension in both directions.
• Synchronized audio (newer versions). Recent LTX models generate coherent audio and video together in one pass — ambient sound, foley, even dialogue — where most open models output silent clips.

The throughline: LTX is built for fast, grounded, natural shots, and it reads prompts as prose.

The One Rule That Matters Most: Write a Paragraph

LTX Video works best with a single flowing paragraph in present tense — not bullet points, not labeled fields, not fragmented instructions. Think of it as a shot description handed to a cinematographer. Include the subject, environment, action, camera behavior, lighting, and style, but weave them together naturally instead of separating them into sections.

Here’s the difference in practice.

What fights the model (fragmented):

Subject: woman, red coat. Setting: city street, night. Action: walking. Camera: tracking. Lighting: neon. Style: cinematic.

What LTX wants (flowing paragraph):

A woman in a red coat walks down a wet city street at night, neon signs reflecting in the puddles around her feet as she moves past a row of shop windows. The camera tracks alongside her at walking pace in a medium shot. Cool blue and magenta neon light catches the rain in the air, cinematic and slightly desaturated.

Same information. The second version is what produces a clean result. If you’ve worked through our universal prompt formula, think of LTX as the model that wants all six elements dissolved into one paragraph rather than stacked as parts.

LTX Video Prompt Examples

Each example is written as a single present-tense paragraph — the format LTX reads best.

1. Everyday Realism

A man in a gray sweater stands at a kitchen counter slicing a tomato on a wooden board, the knife moving in steady even strokes as the slices fall to one side. Morning light comes through the window to his left, soft and warm, catching the steam from a kettle in the background. The camera holds a static medium close-up on his hands with a shallow depth of field, natural and unstyled, like a documentary kitchen scene.

Why it works: A grounded everyday action, written as flowing prose, with the camera and lighting woven in. This is squarely in LTX’s comfort zone.

2. Vertical Social Clip

A young woman in workout clothes laces up her running shoes on a set of concrete steps at dawn, then stands and stretches her arms overhead before jogging off down a quiet tree-lined path. The camera tilts up with her as she rises, then holds a wide vertical shot as she runs away from the lens. Soft pink early-morning light filters through the trees, fresh and clean, with a crisp natural color palette.

Why it works: Written for a 9:16 frame, it keeps the action simple and continuous, which suits real-time generation.

3. Product in Motion

A pair of white wireless earbuds rests in an open charging case on a marble surface, slowly rotating as the case turns on a hidden turntable, the small LED on the case glowing softly. The camera circles in a smooth slow orbit at table height, holding a tight close-up with shallow focus. Bright even studio lighting fills the frame with no harsh shadows, clean and minimal, with a polished commercial finish.

4. Atmospheric Scene

Rain falls steadily on the window of a quiet cafe at night while a single cup of coffee sends up a thin curl of steam on the table beside the glass. The camera drifts in a very slow push toward the window, the city lights beyond softening out of focus. Warm amber light from inside contrasts with the cool blue glow of the street, intimate and cinematic with a gentle film grain.

5. Nature Shot

A small stream runs over smooth gray stones in a forest, clear water folding and splashing around the rocks as fallen leaves drift downstream on the current. The camera follows the water in a slow tracking shot just above the surface, holding a low close-up. Dappled sunlight breaks through the canopy overhead, warm and shifting, with rich natural greens and a realistic, grounded look.

Why it works: Moving water plus a slow tracking shot, all in one continuous paragraph — natural motion described as prose is exactly what LTX renders cleanly.

Tips for Better LTX Video Prompts

Commit to the paragraph format. This is the biggest lever. If you take one thing from this guide, drop the labeled sections and write a flowing present-tense description.

Stay grounded. LTX favors realistic, everyday action. If your idea is highly stylized or surreal, expect more inconsistency — and consider whether a different model suits it better.

Describe camera behavior in the sentence. Instead of a separate “camera: tracking” tag, write “the camera tracks alongside her.” Keep it inside the prose so the model reads it in context.

Use present tense. “She walks,” “the camera drifts,” “light filters through” — present tense matches how LTX was trained and keeps the description active.

Match aspect ratio to platform. Set 9:16 for Shorts and Reels, 16:9 for YouTube and widescreen, before you generate. Write the framing to suit it.

Lean on real-time speed. Because generation is fast, iterate aggressively. Change one element per pass — the lighting, then the camera move — and watch how the output shifts.

Generate LTX-Ready Paragraph Prompts

The paragraph format is simple in theory and surprisingly easy to slip out of, especially if you’re used to tag-style prompting on other tools. LzyPrompt takes your shot idea and returns a single flowing, present-tense paragraph with the subject, action, camera, lighting, and style woven together — the exact format LTX Video reads best. Generate your first prompt free and paste it straight in.

FAQ

Why does LTX Video want a paragraph instead of bullet points?

LTX is trained to read prompts as natural shot descriptions. Fragmented, labeled, or tag-style prompts break that expectation and produce muddier, less coherent output. A single flowing paragraph that covers subject, action, camera, lighting, and style gives the cleanest results.

What tense should LTX Video prompts use?

Present tense. Phrases like “she walks,” “the camera pushes in,” and “light spills across the floor” match how the model was trained and keep the description active and clear.

What is LTX Video best at?

Real-time generation of grounded, everyday motion. It excels at natural, realistic scenes rendered quickly on consumer hardware, which makes fast iteration practical. It’s less suited to highly stylized or fantastical concepts.

Can LTX Video generate audio?

Newer LTX models generate synchronized audio and video in a single pass — ambient sound, foley, and even dialogue — unlike most open-source video models that output silent clips. Check which version you’re running, since this varies by release.

What aspect ratios does LTX Video support?

LTX supports a range including 16:9, 9:16, 1:1, 4:3, 3:4, and 21:9, with durations up to around 10 seconds. Set the ratio to match your target platform before generating and write the framing to suit it.