Blog Posts

How to Write Better Prompts for Runway ML Creative Projects

A vague prompt can leave too much open to interpretation. A focused prompt gives Runway ML a clearer creative brief, which makes it easier to guide the look, movement, and mood of your project.

You don’t need to write a long paragraph every time. You need to describe the parts of the scene that matter most, then refine the result with deliberate changes. The process below helps you move from a rough idea to a prompt you can actually work with.

Start with the main subject and action

Open with the most important information. State what the viewer should see and what that subject is doing.

For example, “a cyclist moving through a misty forest at dawn” gives a stronger starting point than “make a dramatic outdoor video.” The first version identifies a subject, an action, and a setting. Those details give the generation a clear direction.

Use concrete nouns and active verbs. Describe whether a person is walking, turning, lifting, looking toward the camera, or standing still. For an object, explain whether it is rotating, sliding across a table, opening, or catching light.

If the subject is central to the project, place it near the beginning of the prompt. Secondary details can follow.

Build the scene in layers

A useful Runway ML prompt often follows a simple order:

  • Subject and action
  • Location and time of day
  • Composition and camera position
  • Lighting and color
  • Motion and pacing
  • Overall visual direction

This structure keeps the prompt readable and makes revisions easier. If the result feels too dark, you know to adjust the lighting section. If the subject is framed poorly, revise the composition or camera direction without rewriting the entire idea.

Try a prompt such as: “A glass bottle resting on a weathered wooden table, viewed in close-up from a low angle, late afternoon sunlight entering from the left, soft shadows, a slow camera push forward, quiet editorial mood.”

The wording gives the scene a subject, a surface, a viewpoint, a light source, and a movement instruction. It also leaves enough room for creative variation.

Describe the visual language, not just the subject

Two prompts can feature the same subject and produce very different creative directions. Add visual language when the mood or presentation matters.

Useful details include:

  • Close-up, medium shot, wide shot, or overhead view
  • Soft daylight, hard directional light, or low-key lighting
  • Muted colors, warm tones, cool tones, or high contrast
  • Documentary, cinematic, product-focused, surreal, or editorial treatment
  • Still composition, handheld energy, or smooth controlled movement

Choose descriptions that support your goal. A product concept may need a clean composition and controlled camera movement. A mood piece may benefit from atmospheric light and slower pacing. You don’t need to add every possible style word. Too many competing directions can make the creative brief less focused.

When you need inspiration for a polished object study, a classic item such as the Montblanc Star walker Stainless Steel Roller ball Pen can serve as a useful subject. You could prompt a close-up of the pen on a dark desk, with reflections moving across its metal surface as the camera slowly tracks past it.

Give movement its own instruction

Creative video prompts work better when movement is described directly. Explain what moves, how it moves, and how the camera responds.

Compare these two directions:

“A red umbrella in a city street.”

“A red umbrella moves through a crowded city street in light rain while the camera follows from behind with a gentle handheld motion.”

The second prompt adds subject movement, environmental motion, camera behavior, and pacing. You can make it more precise by adding a starting position or endpoint, such as “the camera begins at street level and rises slightly as the subject reaches a glowing storefront.”

Keep movement physically coherent. If your prompt asks for a fast subject, a slow camera, shifting focus, and several major scene changes at once, the result may become difficult to control. Start with one main movement and add complexity during later passes.

Use reference details with purpose

Specific details help establish identity. They can include the shape of an object, the material of a surface, the architecture of a location, or the time period suggested by the styling.

For example, “a black rollerball pen with polished metal details on cream paper” is more useful than “a nice pen on a desk.” If you want a luxury editorial feel, you might describe “careful studio lighting, a restrained dark palette, and a slow macro-style camera move.”

A second possible reference is the Montblanc Meisterstuck Solitaire Doue P163 Classic Black Gold Rollerball Pen. In a prompt, its contrasting black and gold appearance could guide a refined product scene. Treat the reference as part of the visual brief and describe the qualities you want to preserve.

Write prompts for one creative decision at a time

When a first result misses the mark, avoid changing everything in one revision. Identify the main problem and adjust that part.

If the composition is too wide, ask for a closer framing. If the scene feels flat, revise the light direction or add a clearer foreground and background relationship. If the movement is too energetic, request a slower camera motion and fewer changes in the scene.

Small, controlled revisions make it easier to see which wording affects the result. Save strong prompt versions so you can reuse the structure for another project.

If you’re developing several concepts at once, Jar for AI — Project Memory for ChatGPT, Claude & Gemini may help you keep notes, prompt drafts, and creative decisions together. That can be useful when a project develops across multiple sessions and you want to return to an earlier direction.

A practical Runway ML prompt template

Use this template as a starting point:

“[Subject] [action] in [location and time]. [Camera position or framing]. [Lighting and color direction]. [Movement and pacing]. [Overall mood or visual treatment].”

Example:

“A ceramic cup filled with dark coffee sits beside an open notebook on a sunlit wooden desk in the early morning. Close-up framing from a slightly elevated angle. Warm light enters from the right and creates soft shadows. The camera makes a slow lateral move while steam rises gently from the cup. Calm, refined editorial mood.”

You can shorten this template for simple concepts or expand it when the scene needs more control. Keep the most important information near the beginning and remove details that don’t support the intended result.

Common prompt problems to fix

Too many unrelated ideas

Several locations, moods, or actions can compete for attention. Choose one central concept and build around it.

Style without visual detail

Words such as “beautiful” or “epic” are open to interpretation. Explain what those qualities mean through framing, light, color, texture, or movement.

No camera direction

If the viewpoint matters, say where the camera is and how it moves. Otherwise, the composition may not match your intention.

Trying to solve everything in one prompt

Prompting is easier when you work in passes. Establish the subject and scene first, then refine the visual treatment and motion.

Final checklist before you generate

  • Is the main subject clear?
  • Have you described the primary action?
  • Does the setting support the idea?
  • Have you specified the framing or camera position?
  • Is the lighting direction understandable?
  • Does the requested movement fit the scene?
  • Can you identify the one change to make after reviewing the result?

For a focused starting point, explore Runway ML | Tools for human imagination. Pair the tool with clear creative instructions, review each result carefully, and refine one decision at a time. That approach gives your prompts a stronger purpose and helps your project develop with less guesswork.

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.