Consistent Character: Your Visual Identity with AI

What you will walk away with

By the end of this lesson you will be able to make the same character or aesthetic appear again and again across a series of images — same face, same look, same world — using reference workflows in Midjourney (Character Reference / Omni-Reference) and Nano Banana (reference image in Gemini). You'll produce a 4-image mini-series that looks like it came from one brand, and own a reusable character brief prompt that locks your visual identity in words before you generate a single frame.

Watch the class

Primary video: Midjourney Consistent Character — Wade McMaster

• URL: https://www.youtube.com/watch?v=nyp5KoHiz58
• Duration: ~29 min
• Status: PENDING EMBED VERIFICATION (oEmbed check, embed permission, youtube-nocookie, no regional restrictions — per video-first policy checklist)

Watch it once end to end, then use the manual below to execute at your own pace — the reference method here transfers to any tool even if a button gets renamed.

The concept: consistency is the whole game

A feed of beautiful but unrelated images reads as random. A feed where the same character, palette, and mood repeat reads as a brand — the viewer recognizes you in half a second while scrolling, and that recognition compounds into saves, follows, and the "oh, it's them again" that turns a scroller into a follower.

There are three layers you can hold steady, and you rarely need all three:

• Identity — the same person/mascot's face and body (a recurring host, a brand character). Hardest; use a reference image.
• Style — the same medium and treatment (always flat pastel illustration, always moody film photography). Easiest and often enough on its own.
• World — the same palette, lighting, and framing across everything, even when the subject changes.

Decide which layers matter before you start. The two tools split along these lines (Verified: Jul 2026): Midjourney gives the most control over a consistent character via reference parameters (--cref / Omni-Reference) and is unmatched on art quality — it's paid. Nano Banana (free, inside Gemini) is superb at "keep this subject identical, change the scene" from an uploaded reference — exactly the series-of-posts job. Many creators lock the character in one and vary scenes in the other. Write the character brief once and reuse it every time — words are the cheapest consistency tool you have.

It won't be pixel-perfect: aim for "clearly the same character" and pick the best takes, don't chase an exact clone. The durable skill is reference-driven consistency, not any one syntax; if a button or flag has moved, search the same words.

Step by step: lock one character across a series

You need: one strong "hero" image of your character or a defined aesthetic (make it in lesson 2.1 first), plus a Midjourney subscription (https://www.midjourney.com) and/or a free Gemini account (https://gemini.google.com). Nano Banana alone is enough if you're not subscribing to Midjourney.

Route A — Nano Banana (Gemini), free

1. Open Gemini and click the "+" / image icon to upload your hero image as the reference.
2. Send: "This is my character. Keep this exact person — same face, hair, and outfit. Place them in [NEW SCENE]." Fill in the scene from your Prompt 1 brief.
3. Generate, then keep going conversationally for each new post: "Same character, now sitting at a café, morning light." Re-upload the hero image if the likeness drifts.
4. Download each keeper (image → download icon).

Route B — Midjourney (Character / Omni-Reference), paid

1. In Midjourney, upload your hero image to get its URL (drag it into the prompt bar, or use the reference/image control), or copy the link of an image you already generated.
2. Write your scene prompt, then attach the reference so Midjourney carries the character in — via Character Reference (the --cref <image-url> parameter) or the newer Omni-Reference control, depending on your version. Optionally set how strongly it holds the reference (a character-weight control).
3. Generate several scenes with the same reference attached each time: different backgrounds, same character.
4. Upscale and save your best takes — vary the scene words, keep the reference constant.

Both routes — curate

Lay your outputs out together and keep only the frames where the character is clearly the same. Four strong, consistent frames beat twelve almost-consistent ones.

Prompts (copy and use)

Prompt 1 — Character Brief Builder

Purpose: write your character/aesthetic once, in reusable words, so every generation starts from the same identity instead of your memory. When to use: before you generate anything in this lesson; paste the filled brief into the scene prompts in both routes above. Works in ChatGPT, Claude, Gemini, or Midjourney prompts.

Help me write a reusable CHARACTER BRIEF for a recurring visual identity, then give me
a one-paragraph version I can paste into any image tool.

Fill this in with me:
- WHO: [age range, gender if any, overall vibe — e.g. "late-20s, warm, approachable"]
- FACE & HAIR: [distinct, repeatable features — e.g. "freckles, round glasses, curly
  auburn hair to the shoulders"]
- WARDROBE: [a signature outfit or 2-3 rotating looks]
- PALETTE: [3-4 brand colors that appear in every image]
- STYLE: [medium + treatment — e.g. "clean 3D render" or "warm 35mm film photo"]
- MOOD: [the feeling every image should carry]

Rules:
- Favor 3-4 DISTINCTIVE, repeatable features over many generic ones (distinctive
  features survive across generations; generic ones drift).
- Output two things: (1) the filled brief as a checklist, (2) a single tight paragraph
  I can paste as the character description, with a blank [SCENE] slot at the end.

Example (filled in): WHO = "late-20s, friendly indie-maker" · FACE & HAIR = "round tortoiseshell glasses, short dark curls, light stubble" · WARDROBE = "mustard hoodie" · PALETTE = "mustard, cream, forest green" · STYLE = "soft 3D render" · MOOD = "calm, competent". The paragraph then plugs into every scene: "…in [SCENE: at a standing desk, plants behind]."

Variations: ask for a mascot/non-human version (product, animal, abstract logo character); ask for an ES/PT version of the paragraph; ask for "three signature outfits" to add variety without breaking identity.

Warnings: don't build a "character" that's a real person's likeness without their consent — that's both an ethics and a disclosure problem (Module M8 covers when a realistic depiction of a real person requires a label). Keep your brief; it's a brand asset you'll reuse for months.

Common errors and how to fix them

Exercise

Build a 4-image mini-series with one consistent character or aesthetic:

1. Run Prompt 1 to write your character brief and get the paste-ready paragraph.
2. Generate a hero frame (Route A or B).
3. Generate three more frames in different scenes, keeping the character/reference constant — vary setting, pose, and light only.
4. Lay all four out together and curate: replace any frame where the identity drifted.

Expected result: four images that a stranger would immediately read as "the same character / same brand" — shared face or style, shared palette, shared mood — plus your saved character brief (a reusable brand asset). Keep this series; one of these frames becomes the still you animate in lesson 3.2.

Final checklist

• Wrote a reusable character brief with 3-4 distinctive features (Prompt 1)
• Generated a hero frame you're happy with
• Made 3 more frames keeping the character/reference constant
• Held the STYLE + PALETTE steady across every frame, not just the first
• Curated: rejected any frame where identity or style drifted
• Four frames read as one brand at a glance
• Character is not a real person's likeness used without consent
• Saved the brief for reuse

Toolkit for this lesson

Required

Also useful

Model names, parameters, and prices change fast — dated benchmarks, not permanent facts.

Next step

Lesson 2.3 — "Text in Image, Covers & Carousels": your character and style are set; now you'll make images that carry legible text — thumbnails, covers, and a 5-slide carousel — using tools built to spell words correctly, so your brand reads as clearly as it looks.