Midjourney Prompts for People and Personas (2026 Guide)

Midjourney prompts for people work best as four blocks in a fixed order: who the person is, what they are doing and where, how the photo is shot, then parameters such as --ar 4:5 --raw. On V8.2, Midjourney's default model since July 24, 2026, the face itself is carried by up to four reference images attached through the Edit Model, and the look of the feed by a style reference or sref code.

Versions, parameters, plan prices and rules checked October 2026 against Midjourney's docs: versions, parameter list, prompt basics, Edit Model, Style Reference, Raw, plans and community guidelines.

This page used to be a general parameter reference. It is now built around the job most readers arrive with: prompting a believable person, then prompting the same person again tomorrow. It is the prompt-writing part of our AI image generation for influencers guide, which covers choosing the tool in the first place.

One correction for anyone who learned on V7: Omni Reference (--oref) is now a V7-only feature. Midjourney's version docs say the V8.2 Edit Model "replaces Omni Reference, Character Reference, and the Retexture tool", so most character tutorials written before July 2026 teach syntax the default model no longer uses.

How Midjourney prompts work for people

A people prompt is a short description followed by parameters, and the order of the description matters less than keeping it the same every time. Midjourney's Prompt Basics page lists seven things worth naming: subject, medium, environment, lighting, color, mood and composition. For a persona, those collapse into three blocks of text plus the parameters.

Put together, a full prompt reads like this:

27-year-old woman, shoulder-length dark brown wavy hair, slim build, round tortoiseshell glasses, small gold hoop earrings, reading at a bookshop window seat, oversized olive jacket, candid photo, medium shot, overcast afternoon light --ar 4:5 --raw

Three habits from the docs save GPU time. They sound basic, and they fix most first-grid disappointments:

• Keep it short and specific. Midjourney says short, simple prompts typically generate the best images and that long lists of instructions can confuse the model. Name the details that matter and stop.
• Describe what you want, not what you don't. The docs' own example: a party with "no cake" can still produce a cake. Exclusions go in --no.
• Format parameters exactly. They go last, with a space before each -- and no commas or periods inside them.

V8 rewards a little more detail than older versions did. Midjourney's V8 announcement (March 17, 2026) says the model is "much better at following detailed directions", so if the first grid gets the jacket or the light wrong, add that detail rather than starting again.

Midjourney character prompts: the persona block

A Midjourney character prompt starts with a persona block: one fixed line describing the person, pasted word for word at the front of every prompt. It is the text half of consistency. The reference images do most of the work on the face, but the Omni Reference docs note that text is "just as important" for carrying details, and a block you reword every day gives the model a slightly different person every day.

Write the block once, save it somewhere you copy from, and treat changes like a rebrand. If you would rather fill in a form than write it from scratch, our free AI influencer prompt generator turns a persona brief into a block in this shape, and the AI influencer prompt pack has scene ideas sorted by shoot type to put after it.

Midjourney realistic person prompt: getting a photo, not a render

For a realistic person, add --raw, name a photographic medium, describe a real light source and leave --s at 100 or lower. Midjourney's Raw docs describe standard mode as an auto-pilot that adds its own creative touch, and say that with simple prompts Raw often gives "more realistic, photo-like images". The V8 announcement makes the same call: for photos or a more controlled look, switch to --raw right away.

Stylize is the second lever. A low value follows the prompt closely; a high value gives Midjourney more freedom and can drift from your details. For a persona, drift is the enemy, so start at the default and only raise it when a shot is meant to look editorial.

The right-hand column is craft advice, not a Midjourney rule, and it comes from how photographs actually look: lit by something, taken from somewhere, with a subject who is busy. When a grid looks synthetic, the fix is usually to delete adjectives and add one concrete noun.

Midjourney portrait prompts: eight templates for a persona

These eight Midjourney portrait prompts cover the shots a persona account uses most. They are starting points for V8.2, not promised results. Replace [persona] with your persona block, run one grid, then change one element at a time. Use a tall frame such as --ar 4:5 or 3:4 for feed portraits and 9:16 for vertical covers.

1. Hero headshot (your first reference image)

studio headshot of [persona], neutral expression, facing camera, soft even light, plain light gray background, sharp focus on the eyes --ar 1:1 --raw --s 50

2. Three-quarter portrait

three-quarter portrait photo of [persona], looking just past the camera, window light from the left, plain wall behind, 85mm --ar 4:5 --raw

3. Full-body outfit shot

full-body photo of [persona] standing on a quiet side street, [outfit], overcast daylight, shot from waist height, whole figure in frame with space above the head --ar 3:4 --raw

4. Candid lifestyle

candid photo of [persona] laughing at a street food stall at dusk, warm string lights, shallow depth of field, natural skin texture, 50mm --ar 3:4 --raw

5. At home, with a fixed look

[persona] making coffee in a small sunlit kitchen, morning light through linen curtains, relaxed mood, medium shot --ar 3:4 --sref [your style code] --sw 150

6. Environmental portrait in the niche

environmental portrait of [persona] at [the place the niche happens], [tool of the niche] in hand, practical light from the scene, medium-wide shot --ar 4:5 --raw

7. Golden hour close-up

close-up photo of [persona] on a rooftop at golden hour, wind in the hair, low sun from behind, soft lens flare, calm expression --ar 4:5 --raw

8. Vertical cover frame

[persona] walking toward the camera through a covered market, motion blur in the background, eye contact, vertical frame with clear space at the top for a caption --ar 9:16 --raw

To explore before you commit to a look, add --c 40 --sref random to any of them: chaos spreads the four results apart and --sref random applies a random style code that you can keep if you like it.

Animating a portrait. Midjourney's video model uses an image as the first frame. Clips start at 5 seconds and extend 4 seconds at a time up to 21. Keep the prompt about motion ("she turns toward the camera and smiles") and leave --motion low, the default, for subtle movement; the docs warn that high motion can look glitchy. Video is 480p, or 720p on Standard and above. For other engines, see our Kling image-to-video tutorial and the list of free AI video generators.

Midjourney parameters that matter for people on V8.1 and V8.2

Fifteen parameters cover almost every people prompt, and four of them (--ar, --raw, --s, --sref) do most of the work. Values come from Midjourney's parameter pages and the V8 column of its version compatibility chart, checked October 2026.

Sources: Aspect Ratio, Stylize, Chaos, No, Seeds, Image Prompts, Personalization and Draft mode. A note on seeds, because the myth persists: Midjourney says a seed cannot capture a character or appearance across different prompts. It is a testing tool, not an identity tool.

Parameters that no longer work on V8

• --oref and --ow: Omni Reference is V7-only. The parameter list marks it "replaced by the Edit Model in V8.X".
• --cref and --cw: Character Reference is a legacy feature from the V6 era.
• --q (quality): V6 accepted 0.5, 1 and 2, V7 accepts 1, 2 and 4, and the V8 column of the chart lists no values.
• Multi-prompts and weights (::): supported only up to V6.1 and Niji 6 (Multi-Prompts).
• --turbo: not supported on V8.1 or V8.2 (GPU Speed).

Keeping the same person: the Edit Model replaced Omni Reference

On V8.1 and V8.2, you keep a person consistent by attaching up to four reference images to the prompt through the Edit Model. It does three jobs: it changes an existing image from written instructions, it generates new images from references, and it powers inpainting and outpainting in the Editor. It works with style references, moodboards, Personalization and --hd.

If you prefer the V7 look, the older syntax still runs there. Add --v 7, then --oref with one image URL and --ow for strength:

[persona] reading at a bookshop window seat, candid photo, overcast light --ar 4:5 --raw --v 7 --oref https://your-image-url.png --ow 100

The Omni Reference docs set the limits: one image only, --ow from 1 to 1,000 with a default of 100, a recommendation to stay below 400 unless stylize is very high, and twice the GPU cost of a normal V7 job. Our Midjourney consistent character workflow walks through both versions shot by shot, and the consistent character AI guide compares Midjourney with every other method.

Style codes: one look for the whole feed

A style reference gives every image the same look without touching the person in it. Midjourney's Style Reference docs say it captures colors, medium, textures and lighting and "doesn't copy objects or people". You can point --sref at an image, or at a numeric style code from Midjourney's own library, which is easier to reuse across a whole account. If you would rather have a reference photo's look in words you can edit, upload it to our free image to prompt generator, which writes a prompt in Midjourney format.

• Finding a code. Open the Explore page and click the Styles tab to reach the Style Explorer. Search a word such as "photographic", click a code to see examples, and like the ones you want to keep.
• Rolling the dice. --sref random applies a random code, and the prompt then shows the number so you can reuse it.
• Setting strength. --sw runs from 0 to 1000 with a default of 100. It does not work with moodboards.
• Mixing. You can use more than one code, or combine codes with images. You cannot turn an uploaded image into a code.
• Keeping the text about content. Midjourney's example of a good prompt is "detailed portrait of a dog"; its example of a bad one is "copy this style and make a bunny".

Personalization (--p) is the other look tool. It learns your taste from images you select, and your V7 global profile works on V8.1 and V8.2. Stylize controls how strongly a profile is applied, which is one more reason to leave --s alone once a persona's look is settled.

Troubleshooting people prompts

Most people-prompt failures trace back to a missing reference, default styling or a parameter the current version no longer reads. Find the symptom, apply the fix, rerun one grid.

What a persona month costs in Midjourney

Midjourney has four subscription tiers and no free plan in its plan comparison: Basic at $10 a month, Standard at $30, Pro at $60 and Mega at $120, with 20% off when billed annually (checked October 2026). The free trials page has the current trial status. The practical difference between tiers is Fast GPU time:

Midjourney lists an SD job at 0.8 minutes of GPU time, an HD job at 1.3 minutes, and a V8 Draft batch of 24 small images at 0.4 minutes. As arithmetic on those figures, Basic's 200 Fast minutes is 250 SD jobs or about 153 HD jobs a month. Standard adds unlimited image generation in Relax mode, which is why it is the usual floor for a daily-posting persona. Extra Fast time is $4 an hour on every plan.

Rules for prompting people in Midjourney

• Safe for work only. The Community Guidelines ban adult content and gore, and they apply in Stealth mode and direct messages too.
• Uploads. You must have the rights to any image you upload as a reference.
• No misleading people about the source. The guidelines say not to intentionally mislead recipients of generated images about their nature or source. Label a persona as AI where a platform asks for it; our AI Instagram model guide covers Instagram's labels.
• Visibility. Images are public by default. Stealth mode, on Pro and Mega only, keeps them off the website.
• Commercial use. The plan page says subscribers can use their images in "just about any way", with one stated condition: a company making more than $1,000,000 a year in gross revenue must be on Pro or Mega. Read the Terms of Service for the detail.
• No automation. Midjourney says it does not provide an API, with rare exceptions it grants explicitly, and the guidelines prohibit automating the service or using third-party apps.

When Midjourney is the wrong tool for a persona

Midjourney is the right tool when you want strong default aesthetics, style codes and a hosted app, and you are comfortable with close-but-not-identical faces. It is the wrong tool in three cases.

• You need the exact same face across hundreds of images. References are quick but not exact. A character LoRA on an open model is the dependable route; see our LoRA training guide and what ComfyUI is.
• You need automation. A pipeline that generates on a schedule needs a model with an API or a local install. Our Midjourney vs Flux comparison and Flux vs SDXL cover the open options.
• You want to put a real face on generated footage. That is a different tool and a different set of consent rules; start with the AI face swap guide.

Prompting is the first skill, not the whole job. Our AI Influencers program picks up where this page stops: its Character Design module covers creating the persona, training a face LoRA and body consistency with ControlNet, and its Content Production module covers photorealism and AI video.

Midjourney prompts FAQ