Luma Dream Machine Prompt Guide: Create Cinematic AI Video

Luma Dream Machine went from a scrappy beta in 2024 to one of the most reliable AI video generators on the market. Its physics engine is genuinely impressive — water flows correctly, fabric drapes naturally, objects fall with real weight. But most people underuse it because they write prompts the wrong way.

The difference between a muddy, confusing clip and a cinematic one almost always comes down to how you structure the prompt. This luma dream machine prompt guide covers what the model actually responds to, how to control the camera, and how to use features like @style and End Frame that most people skip entirely.

How Luma Dream Machine Handles Prompts

Luma doesn’t want a novel. It doesn’t want a keyword dump either. The sweet spot is 20 to 40 words — roughly 3 to 4 sentences that describe the scene, the action, and the camera. Anything longer and the model starts ignoring details or producing inconsistent results.

This is shorter than what works on Kling or Sora. Luma rewards precision. Every word should earn its place in the prompt.

The other thing to understand: Luma responds extremely well to cinematic language. Describe the camera as if you’re writing a shot list for a DP, not instructions for an image generator. “Slow dolly forward through the doorway” will produce a better result than “camera moves forward into room.”

If you’re coming from other models, check out the universal prompt formula for the foundational structure. What follows is the Luma-specific version.

The Luma Prompt Structure

After extensive testing, this three-part structure consistently produces the best Dream Machine output:

Scene + Action → Camera → Style

Keep it tight. Three to four sentences total.

Part 1: Scene and Action

Set the location and describe what is happening in one or two sentences. Be specific about environment and motion.

A golden retriever runs along a wet beach at sunset, kicking up sand and water with each stride, waves crashing in the background.

Notice: concrete details (wet beach, sunset, kicking up sand) but no fluff. Luma needs just enough to build the scene.

Part 2: Camera

This is where Luma shines. The model understands standard cinematography terminology and executes it with impressive accuracy.

Camera keywords that work well:

• Pan left / Pan right — horizontal camera rotation
• Zoom in / Zoom out — focal length change
• Dolly in / Dolly out — physical camera movement forward or backward
• Orbit — camera circles the subject
• Drone shot — elevated aerial perspective
• Tracking shot — camera follows the subject’s movement
• Low angle — camera positioned below the subject looking up
• FPV view — first-person perspective, immersive
• Crane shot — vertical camera movement, high to low or low to high

Combine a camera move with a speed modifier: “slow orbit,” “fast tracking shot,” “gentle zoom in.” Luma interprets speed consistently.

Slow tracking shot following the dog from a low angle, shallow depth of field.

Part 3: Style

One sentence for the visual treatment. Film stock references, color palettes, and lighting descriptors all work.

Golden hour light, warm color grading, shot on 16mm film.

The @style Feature

Most people don’t use this and they’re missing out. Luma’s @style feature lets you upload a reference image that defines the visual style of your generation. The model extracts the color palette, lighting quality, texture, and overall mood from that image and applies it to your video.

This is powerful for brand consistency. Upload a frame from your existing content — a color-graded still from a previous shoot, a mood board reference, a film still — and Luma will match that look across every generation.

The prompt still controls content and motion. The @style image controls the visual treatment. Together they give you a level of creative control that text alone can’t achieve.

The End Frame Feature

Luma’s End Frame lets you upload a start image and an end image. The model generates video that morphs between the two, filling in the motion and transition.

This is incredibly useful for:

• Product reveals — start with a closed box, end with the product displayed
• Transformation sequences — day to night, season changes, before and after
• Character movement — start pose to end pose with natural motion between
• Scene transitions — smooth morphs between two different compositions

The key is making sure both frames share enough visual similarity that the model can interpolate between them logically. Two completely unrelated images will produce chaos. Two frames from the same scene with different compositions will produce magic.

5 Luma Dream Machine Prompt Examples

Here are complete prompts you can use directly. Each one stays within the 20-40 word target.

1. Cinematic Portrait

A woman with short silver hair stands on a rooftop at dusk, city lights flickering below. She turns slowly toward the camera. Slow dolly in, medium close-up, shallow depth of field. Moody blue-hour lighting, anamorphic lens flare.

2. Product Shot

A ceramic coffee mug sits on a wooden table, steam rising in soft morning light through a window. Slow orbit around the mug, close-up. Warm tones, minimal, clean commercial aesthetic.

3. Nature and Atmosphere

Dense fog rolls through a redwood forest at dawn, shafts of light breaking through the canopy. Slow drone shot ascending through the trees. Muted greens, film grain, ethereal mood.

4. Action Sequence

A skateboarder launches off a concrete ramp in an empty pool, board rotating beneath their feet. Low angle tracking shot, fast shutter speed. Golden hour, high contrast, documentary style.

5. Abstract and Artistic

Ink drops falling into clear water in slow motion, swirling into organic shapes, deep indigo and crimson expanding outward. Macro lens, extreme close-up, shallow depth of field. Dark background, studio lighting.

Want prompts like these generated for your specific project? Try the LzyPrompt generator — describe your scene and get a ready-to-paste prompt optimized for Dream Machine and other models. It’s free to start.

Tips That Make a Real Difference

Describe the camera, not just the subject. This is the single biggest improvement most people can make. Luma’s physics engine handles subjects well on its own. What it needs from you is clear camera direction.

Keep it under 40 words. Resist the urge to over-describe. If your prompt reads like a paragraph, cut it in half.

Use one camera move per generation. “Dolly forward then pan left then zoom out” produces confused results. Pick one move and commit to it.

Specify speed. “Slow,” “gentle,” “fast,” “rapid” — Luma interprets these consistently. Without a speed modifier, you get the model’s default pacing, which may not match your vision.

Pair @style with every generation. Once you have a visual reference that works, use it on every prompt in your project. Consistency across clips is what separates a collection of AI videos from an actual sequence.

Use End Frame for controlled transitions. When you need specific start and end states, don’t leave it to chance. Upload both frames and let the model handle the in-between.

For short-form content specifically, check out our guide on AI video prompts for TikTok and Reels — many of those techniques pair well with Luma’s strengths.

Frequently Asked Questions

How long should a Luma Dream Machine prompt be?

Aim for 20 to 40 words. Three to four concise sentences covering the scene, action, camera movement, and style. Longer prompts tend to produce less coherent results because the model deprioritizes details at the end.

Does Luma Dream Machine have a character limit?

There is no officially published hard character limit, but testing shows that concise, specific prompts consistently outperform long ones. Treat 40 words as your practical ceiling.

What camera movements does Dream Machine support?

Luma handles standard cinematography terms well: pan, tilt, orbit, dolly, zoom, tracking shot, drone shot, crane shot, low angle, FPV view. Combine these with speed modifiers like “slow” or “fast” for best results.

How do I keep a consistent style across multiple Dream Machine clips?

Use the @style feature. Upload a reference image that defines your color palette, lighting, and mood, then apply it to every generation in your project. This creates visual coherence across clips.

Can I control the start and end of a Dream Machine video?

Yes. The End Frame feature lets you upload a start image and an end image. Luma generates video that transitions naturally between the two. This gives you precise control over the arc of each clip.