Midjourney /blend mode guide
The /blend command fuses 2–5 images into a new one with no text prompt — perfect for merging aesthetics, palettes, or moods when you want the model to focus entirely on the visual qualities of your inputs. Results depend heavily on how many images you use, their visual compatibility, and the dimensions option you select. This guide explains the mechanics and when an image prompt is the better tool.
How /blend behaves with different image counts
The number of images you supply changes the character of the result significantly:
- 2 images: the cleanest and most predictable merge. The model balances traits from both sources. Best starting point for most blend work.
- 3 images: generally still coherent if the sources share a subject, palette, or rendering style. Divergent inputs start to average out.
- 4–5 images: the model averages across all inputs, so fine detail from any one source becomes diluted. Output tends toward a softer, less defined result. Useful for loose mood-mixing or generative texture work, not for carrying specific visual characteristics through.
What /blend does not do
Unlike an image prompt, /blend carries no text component at all and treats its inputs roughly evenly. You cannot bias one image over another using --iw (image weight), because there is no prompt structure to attach that parameter to. If one image needs to dominate, use it as an image prompt with a text prompt and set --iw there instead.
The command also will not infer your aspect ratio from the input images — you must set the dimensions option (portrait, square, or landscape) yourself. Square is the safest default when your sources have mixed orientations.
When to use /blend vs. an image prompt
| Scenario | Better tool |
|---|---|
| Fusing the look of two paintings | /blend |
| Merging a character’s face with a style | /blend with 2 images |
| Combining a reference image with a text description | Image prompt + --iw |
| Keeping one image dominant, adding accents from another | Image prompt + --iw |
| Mixing five mood references | /blend (accept softness) |
Tips for cleaner blends
- Match the orientation of your source images to your chosen dimensions option. A portrait-format source in a square blend often gets cropped awkwardly.
- Keep palettes compatible. Sources with clashing colour schemes — warm sunset vs. cold blue industrial — often produce muddy, desaturated midtones.
- Two strong, visually coherent images consistently beat five weaker ones when you need a defined output.
- Try blending the output. Blend two originals to get a hybrid, then blend that hybrid with a third source for a more layered result than a direct three-way blend.