PromptMake
2026-09-08·16 min read

Runway vs Luma Prompts: Motion Control Differences

Runway vs Luma prompts compared for motion control: Gen-4 dials and Director Mode versus Ray3.2 prose, keyframes, and camera-language dialect.

runway vs luma promptsrunwaylumaray3.2gen-4motion controlvideo prompts

Write video prompts for Sora, Kling, Runway & more

Text-to-video or image-to-video — structured motion language.

Try Video Prompt Generator →

Runway vs Luma prompts is a motion-control dialect fight, not a brand loyalty test. Runway Gen-4 family routes (Gen-4, Gen-4 Turbo, Gen-4.5 where your account shows them) pair short subject-and-camera prose with UI dials, Motion Brush on image-to-video, and Director Mode sliders on Gen-4.5. Luma's current video model is Ray3.2 in the Luma App and API; Dream Machine and Ray2 are deprecated names that still appear in old search. Ray rewards natural cinematic sentences, Multi-Keyframe bridges, and camera phrases in the prompt itself. This head-to-head stays on prompt text and motion language. Soft path to structure either host: https://promptmake.net/video. PromptMake writes prompt text. It does not render Runway or Luma MP4s.

What "runway vs luma prompts" actually compares

You type runway vs luma prompts when the same brief fails on one host and sings on the other. The brief is usually five or ten seconds: product bead, portrait blink, walk approach, slow push-in. Both hosts want subject motion separated from camera motion. They disagree on where control lives. Runway splits control across text, dials, and brushes. Luma puts more of the camera story in prose and keyframe anchors.

As of September 2026, confirm picker labels in your workspace. Runway public help still centers Gen-4 prompting for image-to-video with subject, camera, and scene motion pillars. Gen-4.5 adds Director Mode camera sliders (horizontal pan, vertical pan, zoom, roll, tilt) on supported plans. Luma's official assistant-facing guidance names Ray3.2 as the current video model and treats Dream Machine / Ray2 as legacy. Soft docs and learning-hub Ray2 pages still exist; prefer live App pickers over archived tutorials.

Strong fit for creators who cut the same hero still on both hosts and need a dialect map. Skip this page if you only need a pan keyword sheet that ignores vendor UI. Soft product path when you want structured video prompt text before you spend credits: https://promptmake.net/video.

Motion control on Runway Gen-4 family

Runway Gen-4 prompting guidance favors a foundational motion first, then subject motion, camera motion, scene motion, and style. Official tips push positive language. "Locked camera, subject motion only" travels better than a wall of NO MOVEMENT. Image-to-video already owns wardrobe and lighting. Your text owns delta: timed action and one camera path with speed.

UI partners matter. Camera and motion dials set amplitude. Motion Brush (on Gen-4 image-to-video routes that show it) paints which regions move and in which direction. Director Mode on Gen-4.5 (plan-gated on many accounts) exposes pan, zoom, roll, and tilt sliders roughly from -100 to +100. Keep prompt mood and slider path aligned. "Static shot" in text plus Zoom +50 in Director fights itself. Moderate slider values (often under about ±50 to 70) keep physics calmer than maxed dials. Confirm live UI; labels shift.

Duration is commonly five or ten seconds on Gen-4 family clips. Lead with camera or subject finish lines when order matters. Timestamp style cues help on text-to-video routes that document them. Iterate one pillar per credit: motion wrong, camera wrong, or identity drift.

Runway prompt dialect that survives first credits

Skeleton: [subject action with timing]. [one camera path with speed]. [mild scene motion if needed]. [preserve locks on image-to-video]. Example product bead: "One condensation bead slides down the left side of the bottle between two and four seconds. Locked camera. Soft steam only. Preserve brushed steel finish from the reference." Example push-in: "Slow dolly push from medium to close over six seconds. Subject stands still except one blink at two seconds."

Keep clauses short. One to three sentences beat essay prompts on Gen-4. Refer to people as "the subject" on image-to-video so capacity goes to motion. Match implied pose in the still. A seated portrait wants breath and blink, not a sprint.

Runway UI + text partnership

Use dials when you need the same pan direction across retries. Use text when you need story: what the pan reveals, when the push starts. Use Motion Brush when only the curtain should sway and the face must stay still. Use Director Mode sliders when Gen-4.5 is selected and you need numeric pan/zoom instead of adjectives alone. Soft tip: draft the prose stack on https://promptmake.net/video, then tune dials inside Runway.

Motion control on Luma Ray3.2

Luma controls camera primarily through language and keyframes. Official Dream Machine-era docs said to put camera motion strings in the prompt and listed supported moves via a camera motion API endpoint. Ray3.2 continues the prose-first habit with richer production controls: text-to-video, image-to-video, Modify Video, Reframe, Multi-Keyframe direction (up to sixteen keyframes on supported routes), native HDR / EXR on production tiers, and 5 or 10 second native clips for many t2v/i2v jobs. Confirm which controls your plan shows.

A cinematic motion kit for Ray reads like a short take: preserve or subject card, subject beat with timing, one camera path with speed, duration and aspect. Natural phrasing works: "crane down," "tracking shot," "camera circles slowly," "slow pan upward." Ordered structure still helps: main subject → action → details → scene → style → camera → reinforcer. Vague emotion words waste credits. "His hands shake as he paces" beats "he looks nervous."

Keyframes change the dialect. Start and end frames (and Multi-Keyframe anchors) ask for continuous motion between anchors, not a six-shot montage in one box. Extend and loop features exist on Luma lineages; verify live toggles. Soft path for the text layer: https://promptmake.net/video.

Luma prompt dialect that survives first credits

Skeleton: Maintain [locks] from reference. [Subject beat with finish window]. [One camera path with speed]. [Duration / aspect]. Portrait micro-kit: "Maintain face, hair, and charcoal sweater from reference. Soft smile growth; one blink at two seconds. Camera locked. Five seconds. Vertical 9:16." Walk kit: "Maintain jacket and hair from reference. Walks toward camera at a steady pace and stops in medium close-up by second nine. Slow push-in only. Horizontal 16:9. Ten seconds."

Product turntable: "Preserve matte black bottle and white sweep from upload. Gentle turntable right about thirty degrees. Camera locked. Soft key upper left preserved. Five seconds. No logo invent." Put camera nouns in the sentence. Do not assume Runway-style dials exist on the same surfaces.

Keyframes versus prose-only on Luma

Prose-only wins for simple locked-camera micro-motion. Start-end or Multi-Keyframe wins when the open and close frames disagree on angle or expression and you need a continuous bridge. Write "flowing, no cuts" between anchors. Luma fills light and motion; it will not invent hard editorial cuts you never supplied as separate generations.

Head-to-head: same brief, two dialects

Fair tests use one still, one duration, and one success metric. Run the same job on both hosts before you crown a winner. Soft structure both pastes at https://promptmake.net/video, then adapt dialect. Pricing, credit math, and plan gates change; re-check vendor pages. This section stays on prompt text behavior. Score each take on three checks only: identity hold, motion finish inside the seconds, and camera path match. Ignore style preference until those three pass on both sides. Export both winning prompts into one note so the next shoot reuses the dialect map.

Product bead, locked camera, five seconds

Runway paste: "One condensation bead slides down the left side between two and four seconds. Locked camera; subject motion only. Preserve bottle silhouette and label color blocks from the reference." Set motion dials low or static. Skip Director zoom if you asked for locked camera.

Luma paste: "Maintain brushed steel bottle and white sweep from reference. One condensation bead slides down the left side between two and four seconds. Camera locked. Soft key preserved. Five seconds." No dial partnership required. If the still already shows a bead mid-slide, match that physics.

Expect Runway to obey brush regions when only the bottle surface should animate. Expect Luma to lean on clear preserve + timing sentences. If faces or logos morph, reduce motion amplitude on Runway dials; strengthen preserve language and shorten travel on Luma.

Slow push-in portrait, ten seconds

Runway paste: "Slow dolly push from medium to close over eight seconds. Subject stands still except one blink at three seconds and a soft smile growth. Preserve face and wardrobe from the reference." Pair with a moderate push dial if Gen-4.5 Director or camera controls are active. Avoid stacking orbit language.

Luma paste: "Maintain face, hair, and navy blazer from reference. One blink at three seconds; smile grows once. Slow dolly in from medium to close over the full ten seconds. Locked height. Horizontal 16:9." Optional: start frame medium, end frame close-up, continuous push, no cuts.

Runway often wants dial and text agreement on push strength. Luma often wants the finish frame or a clear end framing clause. Full orbits inside five seconds smear faces on both; prefer short arcs.

Which host for which motion job

Choose Runway when you need region-level Motion Brush, numeric Director Mode camera on Gen-4.5, or a workflow that already lives inside Runway's editor apps. Choose Luma Ray3.2 when you want prose-first cinematic kits, Multi-Keyframe story beats, Modify Video / Reframe pipelines, or HDR/EXR finishing on production tiers. Choose neither as a religion. Soft credit tests beat forum loyalty.

For cross-host pan/dolly keyword sheets, use the camera movement video prompts article on this blog. For Runway-only Gen-4 craft, use the Runway Gen-4 prompts guide. For Luma-only kits, use the Luma Dream Machine prompts guide (Ray3.2 naming). Stay here for the dialect comparison.

PromptMake /video structures text-to-video and image-to-video prompt text for hosts including Sora, Runway, Kling, Pika, Luma, and Veo. Guests get limited daily video-path runs; free accounts raise the cap. Confirm live quotas. Soft start: https://promptmake.net/video.

Common mistakes when porting runway vs luma prompts

Pasting Runway dial assumptions into Luma. Luma will not read a mental Zoom +40. Put the move in words or keyframes.

Pasting Luma essays into Gen-4. Long mood stacks dilute motion. Shorten to action + one camera path.

Contradicting UI and text on Runway. Static prose plus high motion dials invent chaos. Align both channels.

Full orbits in five seconds. Both hosts smear identity. Use short arcs or longer durations.

Re-describing the jacket on image-to-video. The still owns wardrobe. Text owns delta.

Using deprecated Dream Machine / Ray2 strings as current truth. Prefer Ray3.2 pickers as of September 2026. Hedge soft UI renames.

Burning credits on three pillars at once. Fix motion, then camera, then identity.

FAQ

What is the difference between Runway vs Luma prompts for motion?

Runway Gen-4 family splits motion across short prose, UI dials, Motion Brush, and Gen-4.5 Director Mode sliders. Luma Ray3.2 keeps more camera control in natural-language prompts and keyframe anchors. Soft structure either dialect at https://promptmake.net/video.

Is Dream Machine still the right Luma name in 2026?

Search still uses Dream Machine. Official Luma guidance as of mid-2026 names Ray3.2 as the current video model and treats Dream Machine / Ray2 as deprecated. Confirm the live App or API picker before you paste old forum strings.

Does Runway Gen-4 need Director Mode for every shot?

No. Many Gen-4 jobs work with prompt text plus basic camera dials. Director Mode on Gen-4.5 helps when you want numeric pan/zoom/roll/tilt. Plan gates apply on many accounts. Keep text and sliders aligned.

Can I use the same prompt on both hosts?

Start from one fact stack (subject beat, camera path, duration, preserve locks), then translate. Shorten for Runway. Expand camera nouns and keyframe notes for Luma. Expect to retune after one credit each.

Does PromptMake render Runway or Luma video?

No. PromptMake generates structured prompt text on /video (and aliases). You paste into Runway or Luma to render. Soft start: https://promptmake.net/video.

Which is better for product micro-motion?

Both work when you lock the camera and time one bead or fabric settle. Runway helps if you need Motion Brush on a region. Luma helps if preserve sentences and a clean still already tell the story. Test one still on both before you standardize.

Where do Multi-Keyframe and Motion Brush fit?

Motion Brush is a Runway image-to-video control for painted region motion. Multi-Keyframe is a Luma Ray control surface for anchoring looks and beats across a clip. Neither replaces clear subject-and-camera prose. Use UI tools to reinforce the written beat, not to fight it.

Ready to generate your own prompts?

Free. No sign-up required. Works with all major AI models.

Related articles