By the How To Write Prompt Team · Last updated: November 12, 2026 · 11-minute read
Quick answer: To get a three-panel yoga collage from an AI image model, open your prompt with the layout ("a cinematic three-panel vertical collage"), describe each panel's pose separately, repeat the same outfit and lighting in every panel, and use a reference image for the face. Copy-paste prompts for Midjourney and Flux/DALL-E 3 are below.
Introduction: Why One Yoga Image Is Easy and Three Are Hard
Asking an AI model for one picture of a woman doing yoga is the easy part. Almost every tool can do it in seconds.
The real challenge starts when you want a three-panel vertical collage where the same woman does three different poses, in the same outfit, under the same light. Models want to draw one subject, centered, in one frame. You have to convince them to behave like a magazine designer instead.
In this guide we audit a wellness photography prompt line by line. You will learn how to split a 9:16 canvas into sequential panels, keep the character consistent, and get editorial-quality lighting that suits Pinterest, Instagram Reels, carousels and fitness app branding.
A quick honest note before we start: AI tools update often, and results change with model version, settings and luck. Treat every prompt here as a strong starting point and expect to tweak a word or two.
What You Will Learn
- How to write the layout instruction so the model draws three panels
- How to describe each pose so the panels feel connected
- Why a deep burgundy outfit and soft diffused light work so well
- Ready-to-use prompts for Midjourney v7 and DALL-E 3 / Flux.1
- Five extra collage variations you can copy today
- The most common mistakes and how to fix them
How We Approach Testing
We follow a simple rule at How To Write Prompt: tested, not guessed.
- We start with the goal, not the keywords. What is this image for?
- We write the layout first, then the subject, then the light.
- We rewrite the same idea for each platform, because Midjourney and Flux read prompts differently.
- We run the prompt several times. A prompt that works once may be luck.
- We cut every word that does not change the picture.
(Publisher tip: add your own real before-and-after screenshots here, with the date and tool version. First-hand evidence is the strongest trust signal you can give readers and search engines.)
Understanding the Editorial Design Objective
The first rule of good prompting is to name the true goal. The goal of this prompt is not "draw a girl doing yoga." It is to produce an Editorial Wellness Asset.
Wellness brands lean on multi-image storytelling to show a routine, a flow or a transformation. A vertical three-panel collage looks like a page from a high-end fitness magazine or a swipeable social carousel. To work, it needs three things:
- Serenity: a calm, uncluttered scene
- Clarity: clean body lines and readable poses
- Hyper-realism: natural skin, believable fabric, real-looking light
Who is this useful for? Yoga instructors making class promos, wellness brands designing ads, app designers building onboarding screens, and creators making Pinterest pins.
Prompt Audit: Scoring the Collage Prompt
Here is the foundation we are auditing. Each part of the prompt has a job.
Prompt element
What it does
Why it matters
"A cinematic three-panel vertical collage"
Sets the layout
Overrides the model's habit of drawing one centered subject
"Same woman, same outfit"
Locks the character
Keeps all three panels looking like one photoshoot
Deep burgundy outfit
Colour anchor
Creates a strong focal point in every panel
Calm, minimal studio
Controls the scene
Removes distractions and keeps the mood peaceful
Soft diffused lighting
Controls the mood
Avoids harsh shadows that break the calm feeling
Shallow depth of field
Directs attention
Blurs the background so the pose is the star
9:16 aspect ratio
Sets the format
Matches Reels, Stories and Pinterest
Face matches reference photo
Identity control
Keeps the same face across all panels
If we were scoring it, the layout line and the repeated outfit/lighting details carry the most weight. Without them the collage usually falls apart.
Composition Engineering: Hacking the Grid Layout
Most beginners struggle to get several images in one generation. The secret is composition engineering.
By opening the prompt with "A cinematic three-panel vertical collage," you tell the model how to divide the canvas before it decides what to draw. If you bury that phrase at the end, the model has already committed to a single-frame composition.
Then describe each panel on its own, in order, from top to bottom. For this guide we use the following sequence:
Panel
Pose
Shape it creates
Top
Mountain Pose with arms raised overhead
Tall
Middle
Warrior II
Wide
Bottom
Child's Pose
Grounded and compact
This gives a lovely visual rhythm: tall, then wide, then grounded. The eye travels down the 9:16 frame like a slow breath out. (The original source material describes the pattern as tall, wide, grounded. These three poses are our recommended way to get it, and you can swap in your own.)
[Image idea: A vertical three-panel yoga collage showing Mountain Pose, Warrior II and Child's Pose in a burgundy outfit in a minimal studio.] Alt text: Three-panel vertical yoga collage in deep burgundy outfit with soft studio lighting
Visual Psychology and Camera Engineering
Great wellness images are not just about the pose. They are about how the picture feels.
The Outfit: Deep Burgundy
Colour is communication. In wellness imagery, deep burgundy suggests grounding, warmth and earthiness. Against a calm, minimal studio, it becomes a high-contrast anchor. When the viewer's eye jumps from panel to panel, the burgundy ties all three together.
Tip: always repeat the colour name in every panel description. If you only mention it once, models sometimes shift to a different shade or a different outfit by the third panel.
Camera and Lighting: Soft Diffused Light
To get that editorial wellness look, the prompt uses soft diffused lighting and shallow depth of field.
- Soft diffused light removes harsh shadows and creates a peaceful, relaxed mood. Think of window light passing through a thin white curtain.
- Shallow depth of field gently blurs the background, so the viewer focuses on the pose and on realistic skin texture.
Agency-Level Premium Prompts (Copy and Paste)
Here are the ready-to-use prompts. Every prompt is wrapped in curly braces so you can copy it cleanly. Remove the braces if your tool prefers plain text.
1. Midjourney v7 Optimized Prompt
Pro-tip for keeping the same face: Midjourney's reference tools changed with newer versions. In v6 and earlier, character reference used --cref [URL] --cw 100. In v7, use the omni-reference feature instead: --oref [URL_TO_YOUR_IMAGE] --ow 100. Check Midjourney's current documentation, because these parameters can change.
2. DALL-E 3 / Flux.1 Translation
These models respond better to full sentences, so the same idea is written as a short description.
Note on face consistency: DALL-E 3 and Flux.1 do not use Midjourney-style parameters. If you need the exact same face as a photo, use a tool or workflow that accepts a reference image (many Flux-based editing tools do), or generate one panel first and ask for the other poses based on it. Always make sure you have the right to use any reference photo.
3. Negative Prompt (Optional)
In Midjourney, add --no followed by what you want to avoid. For other tools, add it as a plain sentence.
Five More Collage Variations You Can Copy
New in this guide. Each variation keeps the same structure (layout first, same outfit, same light) and changes only the mood.
Variation A: Warm Sunrise Studio
Midjourney v7
DALL-E 3 / Flux.1
Variation B: Terracotta Earth Tones
Midjourney v7
DALL-E 3 / Flux.1
Variation C: Cozy Home Living Room
Midjourney v7
.webp)
DALL-E 3 / Flux.1
Variation D: Outdoor Garden Deck
Midjourney v7
DALL-E 3 / Flux.1
Variation E: Evening Wind-Down
Midjourney v7
DALL-E 3 / Flux.1
[Image idea: A grid showing the five variations side by side: sunrise, terracotta, living room, garden deck and evening wind-down.] Alt text: Five yoga collage variations with the same burgundy outfit in different settings
Platform Tips at a Glance
Platform
Best for
What to remember
Midjourney v7
Polished, photographic look
Put the layout phrase first, and use --oref for a reference face
Flux.1
Natural-language prompts and realism
Write full sentences and describe each panel clearly
DALL-E 3
Easy, conversational prompting
Repeat the outfit and lighting in each panel; check the result for layout drift
Common Mistakes When Writing Collage Prompts
- Burying the layout instruction. If "three-panel collage" is at the end, the model may ignore it. Put it first.
- Describing the outfit only once. Repeat it, or write "the same outfit in every panel."
- Using vague poses. "Doing yoga" gives you random shapes. Name each pose.
- Piling on adjectives. "Beautiful, stunning, amazing" does not change the image. Specific details do.
- Mixing lighting styles. Do not ask for soft studio light and harsh midday sun in the same prompt.
- Forgetting the aspect ratio. Without 9:16, the panels can come out squashed or cropped.
- Asking for too many panels. Three is the sweet spot. Four or more often leads to messy results.
- Using someone else's photo without permission. See the ethics section below.
Quick Fixes
- Panels merge into one scene: add "clearly separated panels with thin white borders."
- Outfit changes colour: add "the exact same deep burgundy outfit in all three panels."
- Poses look wrong: describe the body positions simply (arms, legs, head) instead of only naming the pose.
- Skin looks plastic: add "natural skin texture" and lower the stylize value.
Responsible and Ethical Use
- Use reference photos you own or have permission to use. Do not recreate a real person's face without consent.
- Disclose AI-generated imagery where a platform or advertising rule requires it.
- Do not claim AI images are real client results. For fitness or health marketing, this matters a lot.
- Show safe, realistic poses. If you teach yoga, remind your audience that real practice should be guided by a qualified instructor.
- Check each tool's terms for commercial use before publishing.
Frequently Asked Questions
Can AI really make a three-panel collage in one go?
Yes, most of the time. Start the prompt with the layout instruction and describe each panel separately. Expect a few attempts before you get a clean result.
How do I keep the same face in all three panels?
In Midjourney v7, use a reference image with --oref and the --ow weight. In other tools, use an image-reference feature or generate one panel first and build the others from it.
Why a deep burgundy outfit?
It gives strong contrast against a minimal studio and ties the panels together visually. You can swap in any single colour, as long as you repeat it in every panel.
What aspect ratio should I use?
9:16 is the standard vertical format for Instagram Reels, Stories, TikTok and Pinterest.
Can I use these prompts for commercial work?
The prompts are free to use. Whether you can use the generated images commercially depends on the tool's terms, so check them first.
Can I change the poses?
Absolutely. Keep the structure (layout, same outfit, same light) and swap the pose names. Try to keep a mix of tall, wide and low shapes for a good visual rhythm.
Final Prompt Engineering Lessons
Great prompts are not built by adding more adjectives. They are built by combining three things: structural composition (the grid), visual psychology (colour and light), and environmental storytelling (the studio, the mood).
Treat the canvas like a blank magazine page instead of a single photograph, and the AI starts acting like a graphic design engine. Start with the Midjourney prompt above, change one thing at a time, and keep notes on what works. That habit alone will make you a better prompt writer.
About the Author
Written by the How To Write Prompt team, prompt engineers and creators who test prompts across Midjourney, Flux and ChatGPT before publishing them. Our rule: tested, not guessed.
Disclaimer: AI tools change often, and output varies by model version, account and settings. Always review generated images before publishing.