Mira Video Prompt Guide: Better Cinematic Prompts with Shot Design and Clear Motion Direction

Overview

This Mira video prompt guide turns scattered prompt advice into a practical workflow you can actually reuse. The structure below focuses on production-ready video prompting for creators who want cleaner motion, clearer storytelling, and more stable cinematic results.

The goal is simple: stop writing vague cinematic adjectives and start writing prompts like a lightweight shot plan. Mira performs better when you describe subject, motion, scene, camera behavior, style, and constraints as a clear system.

Use Mira as the Base Model

All examples in this guide assume you are using the standard mira video workflow in Miraga. If you have experience with other video generation systems such as Seedance or Veo, you can apply the same prompting habits here, but the examples below are written specifically for Mira usage inside Miraga.

Core Formula

subject + action details + scene environment + lighting and color + camera movement + visual style + quality + constraints

That formula works because video prompting has both a spatial layer and a temporal layer. You are not only telling Mira what the frame should contain, but also how events evolve over time.

Prompting Principles

• Lock the subject first: identify exactly who or what the viewer should follow.
• Describe motion physically: specify body parts, speed, force, and direction instead of saying “more dynamic”.
• Give the camera one job: choose one main move per shot such as slow push-in, handheld follow, or locked tripod.
• Use constraints aggressively: say no subtitles, no watermark, no logo, no extra characters when those details matter.
• Prefer short shot sequences over exact seconds: “Shot 1 / Shot 2 / Shot 3” is usually more stable than trying to micromanage timestamps.

Write Like a Storyboard

For anything more complex than a simple loop, write the prompt as a mini storyboard:

Shot 1: medium shot, a young woman pauses in a neon alley and looks over her shoulder.
Shot 2: handheld follow shot as she starts walking faster, rain reflecting cyan and magenta light.
Shot 3: close-up as she stops beneath a sign, exhales slowly, and the camera pushes in.

This structure gives Mira a sequence of visual instructions instead of one overloaded sentence.

Reference and Edit Patterns

• Reference-based generation: describe what to borrow from the input image, such as character identity, clothing, lighting mood, or environment.
• Video editing: say what changes and what must stay unchanged.
• Continuation: explicitly say whether you want to extend forward or continue the existing action naturally.

Reference the character identity from image 1. Generate a cinematic hallway scene where the same character walks toward camera, subtle wind in the hair, cool blue practical lighting, slow dolly backward, no subtitles, no watermark.

Motion Guidance That Works Better

• Better: “She slowly lifts her right hand, fingers slightly tense, then lowers it while turning her head toward the window.”
• Worse: “She does an emotional movement.”
• Better: “He takes two steady steps forward, shoulders tight, then abruptly stops.”
• Worse: “He acts nervous.”

Concrete body movement gives Mira something visible to animate. Abstract emotional labels should be translated into posture, gesture, pacing, and facial behavior.

Camera Language Cheat Sheet

• Slow push-in: good for tension, intimacy, or dramatic reveal.
• Locked wide shot: good for choreography, blocking, and architecture.
• Handheld follow: good for urgency, realism, and pursuit scenes.
• Lateral tracking shot: good for movement across a space without chaotic perspective shifts.

Do not stack too many camera directives into one shot. “Push in while orbiting while panning while tilting” usually creates unstable results.

Prompt Templates

Text to Video

A lone detective in a rain-soaked street at night, coat collar raised, walking slowly toward camera. Wet pavement reflecting red and blue neon. Slow push-in, cinematic contrast, realistic motion, detailed fabric texture, no subtitles, no watermark.

Image to Video

Reference the character from image 1. Generate a subtle performance shot where she blinks, breathes, and turns slightly toward the light. Warm golden-hour window light, shallow depth of field, gentle camera drift, realistic facial motion, no extra people, no text.

Action Scene

Shot 1: wide shot of a masked runner weaving through a crowded market at sunset.
Shot 2: handheld chase shot as fruit crates fall and people jump aside.
Shot 3: medium close-up as the runner glances back, breathing hard, then jumps over a low barrier.
High detail, grounded motion, cinematic realism, no subtitles, no watermark.

Product Video

A premium silver smartwatch rotating slowly on a dark reflective pedestal. Soft studio rim light, controlled highlights on the metal edges, macro product cinematography, smooth turntable motion, luxury ad style, no text, no logo unless specified.

Common Mistakes

• Too many ideas in one line: split the sequence into shots.
• Conflicting instructions: do not request both “calm subtle motion” and “violent fast action” for the same subject.
• Pure adjective stacking: cinematic, epic, beautiful, masterpiece is weaker than describing visible outcomes.
• No constraints: if text, watermark, or extra characters are unacceptable, say so explicitly.

Final Recommendation

When using Mira for video, think less like a copywriter and more like a director writing a compact shot list. Describe the visible event, make the motion physically legible, keep the camera instructions simple, and tighten the output with constraints. That is the fastest path to better consistency inside Miraga.