The gap between AI-generated and real photography has never been thinner. Here's how to close it completely.
There's a moment when you generate an image in Midjourney and you do a genuine double-take. Not because it looks good — you expected that — but because it looks *real*. Like an actual photograph of an actual person. Like someone went out with a Canon R5, set up the lighting, and captured this.
That moment is now the baseline with Midjourney v6.
The sixth version of Midjourney — specifically v6.1, which became the production model in early 2026 — made a significant leap in photorealistic rendering. Skin textures, lighting physics, facial expression, and fine detail all improved dramatically. The result is a model that can produce character portraits that rival professional photography, when you know how to prompt it correctly.
That last part is everything. The model's capability is there — but Midjourney doesn't read your mind. It reads your words. And most people are leaving enormous quality on the table because they're prompting like they're describing a painting rather than briefing a photographer.
This guide is going to fix that. By the end, you'll understand exactly how to build prompts that produce photorealistic character portraits — consistently, intentionally, and at a level that actually impresses people.
First: Understanding What Changed in v6
Before getting into technique, it helps to understand what specifically improved in v6 and v6.1, because those improvements directly shape how you should prompt.
Face and skin rendering became dramatically more accurate. V6.1 significantly reduced the plastic, airbrushed quality that made earlier AI portraits immediately recognizable. Pores, fine lines, subtle asymmetries, the natural variation in skin tone — these render correctly now in a way they simply didn't before.
Lighting physics got smarter. V6 understands how different light sources interact with skin and fabric in a physically accurate way. A directional light source in your prompt doesn't just make the image brighter — it creates realistic shadows, highlights, and the specific quality of light you'd see in a photograph taken with that setup.
Prompt comprehension improved significantly. V6 handles natural, descriptive language better than the keyword-stuffing approach older versions required. You can write a prompt that reads like a photography brief, and Midjourney will interpret it like one.
Hands and anatomy are still Midjourney's hardest problem, but v6.1 reduced errors considerably. If you're composing a tight portrait — head and shoulders, or chest up — you largely sidestep this issue entirely.
The Core Principle: Prompt Like a Photographer
This is the single most important thing to understand about generating photorealistic portraits in Midjourney.
Generic prompts produce generic results. "Portrait of a woman, realistic" will give you something technically competent and visually forgettable. But the moment you start specifying the kinds of details a photographer would think about — the lens, the lighting setup, the distance from the subject, the quality of the light — the output transforms.
A useful mental model: imagine you're briefing a professional photographer before a shoot. You wouldn't just say "take a good picture of someone." You'd describe the mood you're after, the setup, the feel. That's exactly how to talk to Midjourney v6.
The Anatomy of a Great Portrait Prompt
Strong photorealistic portrait prompts in Midjourney v6 typically include six elements. You don't need all six every time, but knowing what each one does gives you precise control.
1. Subject description
Who is this person? Age, gender, distinguishing features, expression, what they're wearing. Be specific but not exhaustive — Midjourney handles well-written descriptions better than long checklists of traits.
Good: "A woman in her early 40s, calm expression, subtle smile, dark hair pulled back loosely, wearing a linen jacket"
Less effective: "beautiful woman, brown hair, brown eyes, pretty, elegant, stunning"
2. Shot type and framing
How close are we to the subject? This shapes the entire composition.
- Close-up / tight headshot: face fills the frame
- Portrait / bust: head and shoulders
- Half-body: waist up
- Environmental portrait: person in context, more background visible
3. Lens and camera
This is where most beginners leave quality behind. Naming a specific lens and camera model isn't just jargon — it tells Midjourney what the image should look like at a technical level. Different lenses produce different compression, bokeh, and depth of field, and Midjourney has learned from enough photography to replicate those differences accurately.
- 85mm f/1.4: the classic portrait lens. Flattering compression, beautiful background blur, subject separation.
- 50mm f/1.8: natural, slightly wider, less background compression. Feels more candid.
- 35mm: wider still, more environmental context, used for street and documentary portraits.
- 135mm f/2: more compression than 85mm, even creamier bokeh, telephoto distance feel.
Camera references that work well: Sony A7IV, Canon EOS R5, Nikon Z8, Leica M11. These give the image a specific color science and image quality profile that Midjourney understands.
4. Lighting setup
Lighting is what separates a snapshot from a portrait. Specify not just the brightness but the quality, direction, and source.
- Natural window light from the left — soft, directional, flattering
- Golden hour sunlight — warm, low-angle, long shadows
- Overcast diffused daylight — flat, even, no harsh shadows
- Rembrandt lighting — one main light creating a triangle of light on the shadowed cheek, classic and dramatic
- Softbox studio lighting — professional, controlled, commercial photography feel
- Practical lighting, warm tungsten — interior light sources visible in frame, moody
5. Texture and detail instructions
These small additions to your prompt significantly affect the realism of skin rendering.
Add: "natural skin texture," "visible pores," "subtle skin imperfections," "light film grain"
These instructions push Midjourney away from the smoothed, digital look and toward something that reads like film or a high-end camera sensor.
6. Parameters
The technical flags at the end of your prompt that control the model's behavior.
- `--v 6.1` — explicitly calls the latest v6 model
- `--style raw` — reduces Midjourney's default stylization, producing more natural, photographic results. This is often the single biggest upgrade for photorealism.
- `--ar 4:5` or `--ar 2:3` — portrait aspect ratios that feel natural for headshots
- `--s 200` to `--s 400` — lower stylize values increase prompt adherence and realism; the default is 100
Ready-to-Use Portrait Prompts (Copy and Adapt)
Here are fully built prompts you can drop directly into Midjourney and modify for your own characters.
The Classic Professional Headshot
> Photorealistic headshot of a woman in her late 30s, confident and warm expression, natural makeup, dark blazer, clean studio background, soft box lighting, sharp focus on eyes, natural skin texture with subtle imperfections, Sony A7IV 85mm f/1.4, --ar 2:3 --v 6.1 --style raw --s 250
The Natural Light Portrait
> Portrait of a man in his mid-40s, slight stubble, thoughtful expression looking off-camera, natural window light from the left creating soft shadows, warm afternoon tones, shallow depth of field, visible skin texture, Leica M11 50mm f/2, light film grain, --ar 4:5 --v 6.1 --style raw
The Dramatic Character Portrait
> Close-up portrait of an elderly woman, deep-set eyes, weathered skin with character, direct gaze into camera, Rembrandt lighting, one side of face in shadow, dark moody background, hyper-detailed skin texture, silver hair, Nikon Z8 85mm f/1.8, --ar 2:3 --v 6.1 --style raw --s 200
The Candid Street Style
> Candid portrait of a young man in his mid-20s, natural unposed expression, street photography style, 35mm lens, ambient city light, slight motion blur in background, natural skin tones, Sony A7IV, photorealistic, --ar 3:4 --v 6.1 --style raw
The Golden Hour Portrait
> Environmental portrait of a woman in her 50s, standing outdoors, golden hour light from behind creating a warm rim light effect, soft bokeh background of blurred trees, natural relaxed expression, Canon EOS R5 85mm f/1.2, warm color grading, visible skin texture, --ar 4:5 --v 6.1 --style raw
---
Advanced Technique: The "Phone Photo" Hack
One of the most counterintuitive techniques for generating photorealistic portraits in Midjourney v6 is starting your prompt with "phone photo of" or "candid smartphone photo of."
It sounds backwards. Why would a lower-quality camera reference produce more realism?
Because smartphone photography has specific visual characteristics — slightly wider lens, deep depth of field, natural exposure, no theatrical lighting — that read as deeply authentic to the human eye. We're trained to see phone photos as real. Midjourney knows this, and outputs generated with this framing tend to look strikingly natural rather than artificially polished.
Try it next to a standard portrait prompt and compare. The difference is often remarkable.
Maintaining Character Consistency Across Images
If you're building a character for a story, a game, a brand, or a content series, you need that character to look like the same person across multiple generations. Midjourney v6 has a feature specifically for this.
Character Reference (`--cref`): Upload an image of your character (or use a previously generated one) and reference it in your next prompt using `--cref [image URL]`. Midjourney will carry the character's facial features into the new generation while following the rest of your prompt for pose, lighting, and context.
Character Weight (`--cw`): Controls how closely Midjourney sticks to the reference. `--cw 100` copies the face very closely. `--cw 50` takes inspiration from it with more variation. For consistent characters, `--cw 75` to `--cw 100` tends to give the best balance of accuracy and adaptability.
This workflow — generate a strong base portrait, then use `--cref` for variations — is how designers and creators are building complete visual identities for fictional characters that maintain consistency across different scenes, lighting conditions, and expressions.
What Still Doesn't Work (And How to Work Around It)
Being honest about Midjourney's limitations saves you frustration.
Hands remain the most common failure point. If hands are visible in the frame, they have a reasonable chance of looking wrong — extra fingers, unnatural bends, strange proportions. The cleanest solution: frame your portraits to avoid hands entirely. Head and shoulders, or a tight crop that cuts off at the chest, sidesteps the issue completely.
Accurate text in images — name tags, signs, writing on clothing — is still unreliable. Use Canva or Photoshop to add any text elements after generation.
Specific real people Midjourney will not generate realistic portraits of identifiable living public figures. This is a deliberate guardrail, not a technical limitation.
Exact likeness repeatability Without `--cref`, generating the same face twice is very difficult. The character reference feature solves this for ongoing projects, but for one-off portraits it takes iteration.
Getting More Out of Your Generations
A few workflow habits that separate people who get consistently great results from those who get great results occasionally:
Generate in batches. Run the same prompt four times (or use Midjourney's 4-image grid) and pick the best. You're not looking for perfection in every generation — you're looking for the one in four that nails it.
Upscale selectively. Only upscale the generations that are genuinely strong. Midjourney's upscaler enhances detail significantly but also reveals any existing flaws at higher resolution. Alternatively, run strong outputs through a dedicated AI upscaler like Magnific AI for commercial-grade resolution.
Vary, don't restart. When you get a generation that's close but not quite right, use Midjourney's Vary (Subtle) option rather than starting a new prompt. It generates variations that preserve the overall composition while adjusting details.
Save your best prompts. When a prompt produces something exceptional, save it. The specific combination of lighting, lens reference, and parameters that worked is worth reusing and building on.
Photorealistic character portraits used to require a photographer, a model, a lighting rig, and a studio. Or years of digital painting skill. Or both.
With Midjourney v6.1, it requires a well-structured prompt and the willingness to iterate until it's right.
The technical ceiling is genuinely there. The question is just whether you're prompting with the precision to reach it.
Share your best Midjourney v6 portrait in the comments — we'd love to see what you're creating. And if there's a specific type of character you're struggling to generate realistically, ask away.
0 Comments