Midjourney Feature Comparison: How to Choose Image Prompts, Style Reference, and Character Reference

In Midjourney, if you want something to look “like a certain image,” “like a certain art style,” or “like the same person,” you need different features. Image Prompts, Style Reference (--sref), and Character Reference (--cref) each emphasize different things—choose the wrong one and your results will drift further off as you tweak. Below, the differences are clarified by goal, along with ready-to-use methods you can apply directly.

What problems do Image Prompts, --sref, and --cref each solve?

Midjourney’s Image Prompt is more like “reference this image’s composition and content cues,” with a more obvious influence on on-screen elements, camera angle, and layout. It’s suited for image-to-image generation, recreating cinematic language, or blending two images into a new composition.

Style Reference --sref focuses more on “borrowing the style,” such as brushwork, color palette, material texture, and overall aesthetics, but it doesn’t guarantee a consistent facial identity. Character Reference --cref is “keeping the same character’s face and key traits as much as possible,” making it better for continuous storyboards, fan characters, or keeping the person consistent across multiple posters.

Control strength: How to combine --iw, --sw, and --cw more reliably

Image Prompts commonly use --iw (image weight) to control how strongly the reference image pulls the result: raise it to look more like the original, lower it for more freedom. Style Reference typically uses --sw (style weight) to adjust how much the style dominates; if it’s too high, it can “overpower” your text prompt and cause the subject to drift.

Character Reference mainly depends on --cw (character weight): higher values look more like the same person, but also more easily bring along fixed expressions or hairstyles. For Midjourney character consistency, it’s recommended to start with a medium --cw as a baseline, then lock clothing, camera, and emotion via text—this is more controllable than simply maxing it out.

Choose by creative goal: The fastest route for three common scenarios

If you want “a set of posters in the same art style,” prioritize Midjourney’s --sref to unify the style, then use text to lock in layout and theme. If you want “the same character in different actions and scenes,” use --cref as the core, specify age, hair color, and signature items in the text, and keep shot framing consistent (e.g., half-body, close-up).

If you want “to replicate the composition of a specific image but swap the subject,” an Image Prompt is the easiest: paste the reference image link, then clearly write what subject to replace it with. If you need both similar composition and similar style, you can use Image Prompt + --sref together, but leave room for the text prompt to work, so the result doesn’t end up only “similar” but not “correct.”

Common pitfalls: Why it seems like you used references but nothing works

The most common issue in Midjourney is that the reference image itself isn’t “clean” enough—for example, the face is too small, the lighting is too busy, or the filter is too heavy—so --cref can’t latch onto stable features. The fix is to switch to a clearer front-facing reference, or first generate a clearer character image using the same reference, then use that image for --cref.

Another pitfall is that your text descriptions contradict each other: you write “realistic ID photo,” but use a strongly stylized --sref, leaving the model torn between directions. When comparing Midjourney features, remember one sentence: Image Prompts handle composition cues, --sref handles aesthetic taste, and --cref handles character identity—the more singular the goal, the more stable the result.