QWE AI Academy Two weeks after you create that perfect blend in Midjourney, the images you uploaded vanish. Not from your gallery – from Discord’s servers. Try to make a variation? Dead link. The blend command’s drag-and-drop convenience has an expiration date.
Most tutorials praise /blend as the “easy” way to merge images. Easy? Sure. What they skip: no text to guide output, duplicate images = account ban risk, Discord’s 25MB cap blocks large files without warning.
The Control Problem: Why /blend Fails When You Need Precision
The blend command analyzes the “concepts and vibes” of 2-5 images. Mashes them together. No text allowed. Quick experiments? Fine. Anything else? Gamble.
One user: “My first experiments were gloopy.” Without text prompts, you’re asking Midjourney to guess from pure visual analysis. Sometimes it nails the aesthetic. Sometimes you get a melted mess.
Discord-only command. On web, use multiple image prompts without text – same result, different path. But uploaded images create temporary Discord links that expire. Old jobs become uneditable ghosts after a few weeks.
When /blend Works (And When It Doesn’t)
/blend shines for visual combinations – mood boarding, style tests, “what if” experiments. On mobile? Lifesaver. Copy-pasting URLs on a phone is miserable. One user noted: “/blend is easier from phone and most devices, without having to get the images url but just drop them easily on discord.”
Skip it when:
- You need text guidance (“make it darker,” “medieval setting”) – use /imagine with image prompts
- You’ll want variations later – uploaded images create links that break
- You’re blending more than 5 images – /blend caps at 5
- Files over 25MB – Discord blocks them
According to official docs, /blend does the same thing as using multiple Image Prompts with /imagine but with a faster interface. Tradeoff: zero text control.
How It Works (Desktop and Mobile)
Open Discord. Type /blend in any channel where Midjourney bot has access. Two image slots appear.
Desktop: Drag files into boxes. Or click to browse. Want more than 2 images? Hit “+4 more” to expand. Select aspect ratio from dimensions dropdown if you don’t want square (1:1 default).
Mobile: Tap image1. Select from camera roll. Repeat for image2. Use “optional” buttons for images 3-5 or dimension changes.
Press Enter. Wait. Blend jobs take longer to start – Discord uploads your images first. Then Midjourney processes. Four variations like any other prompt.
Match your source images’ aspect ratios to your target output. Blending a 2:3 portrait with a 16:9 landscape into a square? Often creates black bars or weird crops. Midjourney inherits framing from the originals.
Three Undocumented Gotchas
1. The Ephemeral Link Trap
Images you upload via /blend live as temporary Discord CDN links. After a couple weeks? They expire. Your blend job stays in your gallery. The “Vary” button? Broken.
One user explained in a WhyTryAI discussion: “The major blend inconvenience is that the images you drop so easily are ephemeral links that will disappear.” If you think you’ll iterate later, host images externally (Imgur, your own site). Use /imagine with permanent URLs instead.
2. The Duplicate Image Ban
Want to weight one image heavier by uploading it twice? Don’t. The Midlibrary guide warns: “Adding the same image more than once is restricted in Midjourney, and trying to bypass it can end in revoking your access.”
No image weight parameter in /blend. Need weight control? Switch to /imagine with the --iw parameter.
3. The Silent 25MB Wall
Discord caps file uploads at 25MB (higher if you have Nitro). Upload a 30MB high-res photo? Nothing happens. No error. It just doesn’t upload. Compress first or you’ll waste time wondering why the slot stays empty.
A Real Test: Portrait + Art Style
I blended a photo of a friend with a screenshot from a Miyazaki film. No text prompt. Result: soft watercolor aesthetic, face structure intact. But clothing morphed into something vaguely medieval. Interesting. Not controllable.
Second attempt with /imagine using the same images plus the text “modern clothing, studio lighting”: clean portrait with Ghibli-style color grading. Clothing stayed recognizable. The text made the difference.
/blend works when you want Midjourney’s interpretation. When you need YOUR interpretation? Add words.
Think of it this way: /blend is fast food. /imagine is the full kitchen. Fast food when you’re hungry and exploring. Kitchen when you’re cooking something specific.
Mobile vs Desktop: What Really Changes
/blend was designed for mobile. On desktop? The benefit shrinks. Dragging files into /imagine isn’t much harder than /blend’s interface. But on a phone, copying image URLs is painful. /blend’s tap-to-select flow saves real time.
One creator noted: “I use the /Blend command when I am out and about on my mobile phone. Image prompts are possible on mobile but very fiddly.” Desktop users gain less. (Remember that ephemeral link problem? Still applies regardless of device.)
Aspect Ratios and Output Quality
Blends default to 1:1 square. Override with 2:3 (portrait) or 3:2 (landscape) via dimensions field. But here’s the quirk: blends “often inherit the framing from one or both initial images” according to Midlibrary analysis.
Translation: source images with different shapes? Midjourney might add black bars or crop unexpectedly even if you specify dimensions. The algorithm tries to preserve compositional elements from both images. Mismatched ratios force compromises.
Crop your source images to match your target ratio before blending. Square, portrait, or landscape – pick one. This gives cleaner data to work with and reduces framing conflicts. Does aspect ratio affect quality? No. But mismatched ratios create composition problems that look like quality issues.
When You Need /imagine Instead
The official documentation states it clearly: “If you want to blend images with a text prompt, use image prompts instead.” That means /imagine.
Format: /imagine [image URL 1] [image URL 2] your text prompt here
This gives you text guidance to steer the blend, image weight control via --iw parameter (0.5 to 2), permanent URLs that don’t expire, ability to blend more than 5 images, and access to all other parameters (stylize, chaos, etc.).
Next Move: Stop Blending Blind
Open Discord right now. Type /blend. Upload two images that have NOTHING in common – a photo of your desk and a Renaissance painting. See what Midjourney does with zero guidance. That’s the tool’s personality.
Then try the same images with /imagine plus “photorealistic office with classical art elements.” Compare results. You’ll immediately see when you need text control and when you don’t.
On mobile and just exploring? /blend is perfect. Building something specific or need to iterate later? Take 20 seconds to use /imagine with permanent image URLs.
FAQ
Can I use /blend with Midjourney V7?
Yes. Works with all current versions including V7 (released April 2025, default since June 17, 2025), V6.1, and Niji models. The command hasn’t changed, though V7’s improved prompt understanding means blends interpret visual concepts more accurately than older versions.
Why do my old /blend jobs show broken images when I try to vary them?
You uploaded images directly through Discord’s /blend interface. Created temporary CDN links that expire after a few weeks. The blend result stays in your gallery, but source images are gone from Discord’s servers. Fix: host images externally (Imgur, Google Drive with public links) and use /imagine with those permanent URLs instead of /blend’s upload slots.
What happens if I blend images with completely different aspect ratios?
Midjourney often inherits framing from one or both source images – creates black bars or awkward crops even if you specify your target aspect ratio in the dimensions field. The blend algorithm tries to preserve compositional elements from both images. Mismatched ratios force compromises. Crop all source images to roughly match your desired output ratio (square, portrait, or landscape) before uploading to /blend.