Setting Up a Structure-Locked Baseline in Midjourney (Before the Curtain vs Lighting vs Rug Test)

Today I was supposed to run a simple comparison test—Curtain vs Lighting vs Rug—to see what actually makes a living room look brighter without changing the layout.
But before I could even start, I hit a few unexpected roadblocks. This post is a quick setup + failure log (and yes, it’s useful), because I finally secured a structure-locked baseline that I’ll use for the next test.

What happened (and why it’s actually useful)

1) My “image” output became a video

I accidentally added my photo through a video workflow (Starting Frame) instead of an Image Prompt workflow.
That’s why the save options showed only raw video / GIF / URL, and I couldn’t export a clean still image the way I expected.
Lesson learned: if you’re doing a layout-preserving interior test, you need to make sure you’re starting in image mode, not video mode.

2) Creation failed: “The sum of all prompt weights must be positive”

I also ran into this error when my prompt formatting got messy—especially when I used --no multiple times or split the prompt across lines.
It’s frustrating in the moment, but it’s actually valuable for a real experiment log: documenting what failed and why increases transparency and trust.

Fix (what worked today)

Fix #1: I adjusted the file size
My image format didn’t change, but when I reduced the file size to 24MB, the upload and generation became stable.

Fix #2: I used --no only once (with a comma list)
Instead of repeating --no, I used it once and listed items with commas. That removed the prompt error and finally generated the baseline.

Baseline prompt (worked):

keep the exact same room layout, architecture, and furniture placement, photorealistic, neutral white balance, natural indoor lighting, no structural changes --no extra windows, ceiling changes, layout change, new furniture

Baseline saved (my “parent” image for the next test)

I saved a structure-safe image as my Baseline.
This isn’t meant to look dramatically different—baseline is a control frame, not a redesign. The goal is to lock the space so I can run clean comparisons later.

Baseline observation notes (quick and practical)

  1. The window and curtain structure stayed consistent (sheer curtain + side curtains), which makes it a reliable reference frame.
  2. The wall-mounted AC unit remained in the same position/scale, which is important for “no-layout-change” tests.
  3. The sunlight and floor shadows still look natural, not overly HDR or “studio staged.”
  4. The overall color tone didn’t change drastically, but it feels slightly more neutral (less warm/yellow cast).
  5. Most importantly, this baseline works as a stable parent frame for the next A/B/C tests—curtain-only, lighting-only, and rug-only.
Image Baseline (structure-locked) 
Baseline (structure-locked) — the fixed reference frame for my next A/B/C tests.
This is the fixed reference frame I'll use for my next A/B/C tests.

Next test: Curtain vs Lighting vs Rug
In the next post, I’ll compare three single-variable changes from this exact baseline:
  • Curtain-onlyTest 
  • Lighting-onlyTest 
  • Rug-only
Key rule: every test must start from the same baseline. If one test builds on another, the comparison stops being reliable.

Transparency note
This post is an AI concept test based on a real room I stayed in while traveling. Results may differ in real life depending on lighting, materials, and camera settings.

Comments

Popular posts from this blog

How I’m Designing a Silent Sanctuary: A Journey from Apartment Noise to an AI-Powered Smart Home

I Tried Midjourney for a Realistic Living Room Redesign (Modern Bright White Test)

What Actually Works in Small Studios: My AI vs Controlled Test Method