Midjourney Series Images Tutorial: Use --sref and --cref to Keep Style and Characters Consistent

To reliably generate a series of images in Midjourney with the same art style and the same character, the key is using “style references” and “character references” the right way. This Midjourney tutorial walks through the workflow in practical order: how to prepare reference images, how to write the parameters, and how to fix common failure cases.

Prepare reference images: clean up the link first

Before using --sref or --cref in Midjourney, prepare 1–3 reference images that are “clear enough.” For style references, prioritize brushwork, color palette, and texture; for character references, prioritize the face, hairstyle, and signature accessories. Ideally, use images with a single subject and no obstruction. Avoid collages and heavy-filtered images, which can distort the defining features.

Upload the images to a Discord chat or the Midjourney web app, then copy the direct image URL/link. Later, pasting the link directly into the same prompt is more stable than sending an image separately and describing it—this is a major turning point for many people when making consistent Midjourney series images.

Lock in an art style with --sref: same “vibe,” even across similar subjects

When writing a prompt, describe the scene normally first, then add “--sref image-link” at the end. Midjourney will prioritize inheriting the style characteristics of that image. You can think of --sref as a “style fingerprint,” ideal for building a set of posters, a consistent e-commerce main image set, or illustrations in the same style.

If the style influence is too strong and the subject turns out “muddy,” make the subject description more specific (clothing, materials, camera, lighting), or reduce the number of style references. If you want it to match the reference style more closely, switch to a “purer” style image (less subject matter, more texture/color blocks)—Midjourney tends to follow that more predictably.

Lock in a character with --cref: the same person across different scenes

When you need “the same character in different scenes,” use “--cref image-link.” This works far better than only writing “the same girl/the same boy,” and it’s especially useful for storyboards, comic strips, and character design sheets. For character reference images, aim for a front-facing view with even lighting. If the face is covered by a hand or the angle is too extreme, Midjourney is very likely to swap the character.

If the character is similar but not close enough, first add more identifiable traits (hair length, bangs shape, mole/scar placement, earrings, uniform details), then gradually fine-tune any character-weight-related parameters (if you’re using them). If you’re generating many consecutive images of the same character, reuse the same prompt structure and key descriptions—Midjourney’s consistency improves noticeably.

Common series issues: why it’s “close, but not consistent”

In Midjourney, “consistent style but inconsistent character” is usually caused by an unclear --cref reference image or a subject description that’s too vague. “Consistent character but drifting style” is often because the --sref style reference isn’t pure enough, or because the prompt mixes conflicting style keywords (for example: watercolor + cyberpunk + photorealistic). Removing conflicting keywords saves more time than repeatedly rerolling.

For series work, it also helps to fix the aspect ratio (for example, keep the same ratio for vertical posters) and keep the camera language consistent (avoid constantly jumping between close-up / half-body / full-body). Once you tighten these variables, Midjourney starts to feel more like a controllable workflow rather than a random slot machine.