Pages

Saturday, September 12, 2026

ChatGPT Image Gen Gets Even Better

Camera ReviewRonMartBlog

ChatGPT Image Gen Gets Even Better

I took two very different reference photos, pasted in a carefully written Viking portrait prompt, and let ChatGPT do the heavy lifting. The results are not perfect, but they show why a good reference image plus a strong prompt is such a useful starting point.

ChatGPT Image Gen Gets Even Better product photo

TL;DR ⚡

ChatGPT can take a reference photo and a detailed prompt and turn it into a surprisingly convincing character portrait. The first result is especially strong; the second is still useful, although I would push it toward a little more grit in a follow-up prompt.

A reference image gets you started. The follow-up prompts are where you dial in the personality.

Table of Contents

  1. Pros and cons
  2. Who it’s for and who should skip
  3. Real World Shots
  4. How I Used the Prompt
  5. Closing thoughts

✅ Pros

  • Strong identity transferThe generated portraits retain recognizable facial characteristics from the reference images.
  • A detailed prompt gives you controlHair, beard, clothing, lighting, composition, and mood are all described before generating.
  • Fast creative iterationThis is a practical way to explore a portrait concept without building the entire image manually.
  • Easy to refineIf the first result is close but not quite there, a follow-up comment can target the weakness.

⛔ Cons

  • The result is not finished just because it looks impressiveThe second portrait still needs a follow-up pass to add the grittier character I wanted.
  • You have to be specificGeneric prompts tend to produce generic character art.
  • Small details can driftGenerated hair, braids, costume details, and facial texture may not follow every instruction perfectly.

Who it’s for

  • Photographers who want to turn an ordinary portrait into a clearly defined character concept.
  • Creators building a visual idea quickly before investing time in a full Photoshop composite or a new photo session.
  • Anyone willing to iterate instead of expecting the first generation to be the final image.

Who should skip

  • Anyone who needs exact product, costume, historical, or facial-detail accuracy on the first attempt.
  • Photographers who are uncomfortable with generative tools doing a substantial part of the image-making process.

Real World Shots 📷

The photos below are real-world samples. Click any photo to open the original size.

I started with an old photograph taken by my friend Robert Vanelli. The goal here was not to preserve the original lighting or hairstyle; it was to give ChatGPT a recognizable face and let the prompt rebuild the subject as a Viking chieftain.

Real-world shot 1
Reference photograph by Robert Vanelli; used as the facial reference for the first portrait.

Click the photo to open the original size 👆

This is the first generated result. I simply supplied the reference image and the prompt, then let ChatGPT do its magic. To my eyes, this is the stronger of the two portraits because the face, dark lighting, fur, and overall character direction come together quickly.

Real-world shot 2
ChatGPT-generated Viking portrait using the supplied reference photograph and the cinematic Viking prompt.

Click the photo to open the original size 👆

For the second example, I used a reference photograph my daughter had taken. I had already sent that image through AI once, so this was another pass through the same general idea rather than a completely untouched starting point.

Real-world shot 3
Reference photograph taken by my daughter; the image had already been processed through AI once.

Click the photo to open the original size 👆

The second final image is a good example of why I treat these generations as a starting point. The overall transformation works, but I wish the result had more grit and lived-in texture. That is exactly the sort of thing I would address in a follow-up comment rather than throwing the whole idea away.

Real-world shot 4
ChatGPT-generated Viking portrait using the second reference image and the same supplied cinematic prompt.

Click the photo to open the original size 👆

How I Used the Prompt

  • Start with the face: I supplied a clear reference photograph and asked ChatGPT to preserve the recognizable facial identity.
  • Separate identity from styling: The reference image provides the face, while the prompt defines the long hair, braided beard, fur, leather, lighting, and environment.
  • State failure conditions: The prompt specifically rules out modern haircuts, short beards, bright commercial lighting, weapons, shields, and a generic Viking actor.
  • Refine instead of restarting: When an image is close but misses the mood, a follow-up comment can add grit, texture, or another targeted adjustment.

Photorealistic Cinematic Viking Portrait Prompt

How to use it: Click Copy Prompt, paste it into ChatGPT, attach a clear reference photo of yourself, and click Send. A sharp photograph with a clearly visible face works best.
Photorealistic Cinematic Viking Portrait Prompt
Create a hyperrealistic, dark cinematic Viking chieftain portrait of the person in the attached reference photograph.

The finished image must look like a real photograph of that exact person physically transformed into a Viking warrior.

This is NOT a generic Viking portrait.

This is NOT a face swap.

This is NOT the reference photograph pasted onto a Viking body.

Reconstruct the subject as if they were genuinely standing in front of the camera wearing the Viking clothing, hairstyle, beard, and fur in the environment described below.


PRIORITY ORDER - FOLLOW STRICTLY

When instructions compete, follow this priority order:

1. Preserve the person's recognizable facial identity.
2. Apply the REQUIRED long Viking hairstyle and REQUIRED long braided Viking beard described below.
3. Preserve the dark, cold, low-key cinematic lighting and tonal treatment.
4. Preserve the tight head-and-shoulders portrait composition.
5. Create realistic fur, leather, and Norse ornamentation.
6. Keep the background extremely subdued and out of focus.

Sacrifice background detail, props, costume complexity, or scenery BEFORE compromising the face, hair, beard, lighting, or composition.


REFERENCE PHOTO RULE

Use the attached photograph ONLY to determine the person's facial identity.

Use it for:

- skull shape
- forehead
- hairline position
- eyebrows
- eyes
- eye spacing
- eyelids
- iris color
- nose anatomy
- cheeks
- lips
- mouth
- chin
- jaw
- ears
- age
- natural facial asymmetry
- distinctive facial characteristics

DO NOT copy the reference photograph's:

- hairstyle
- hair length
- beard length
- facial hair style
- expression
- clothing
- lighting
- background
- pose
- color temperature
- studio look
- skin retouching
- photographic style

VERY IMPORTANT:

The source photograph's existing hair and facial hair must NOT influence the final hairstyle or beard.

Even if the person has short hair in the reference image, the final portrait MUST have the long Viking hairstyle described below.

Even if the person is clean-shaven or has a short beard in the reference image, the final portrait MUST have the long Viking beard described below.

Facial identity comes from the reference photograph.

Hair, beard, wardrobe, lighting, environment, and mood come from this prompt.


FACIAL IDENTITY - EXTREMELY IMPORTANT

The person must remain immediately recognizable.

Preserve the subject's real:

- facial width
- skull proportions
- forehead proportions
- brow shape
- eye shape
- eye spacing
- nose
- cheekbones
- cheek volume
- mouth width
- lip shape
- chin
- jaw
- ears
- age
- facial asymmetry

Do not redesign the face to make it more stereotypically Nordic.

Do not turn the person into a generic handsome Viking actor.

Do not enlarge the jaw.

Do not exaggerate the chin.

Do not change the nose.

Do not alter the eye spacing.

Do not invent dimples.

Do not create artificial facial symmetry.

Do not substantially age the subject.

Someone who knows the person should immediately say:

"That is definitely them."


REQUIRED HAIRSTYLE - DO NOT OMIT

The final subject MUST have long Viking hair.

This hairstyle is mandatory and overrides the hairstyle in the reference photograph.

Create:

- long hair extending approximately to the shoulders or slightly below
- substantial hair volume behind the head
- hair swept backward from the forehead
- forehead clearly visible
- crown hair flowing backward rather than forward
- no bangs
- no fringe
- no modern short haircut
- no undercut
- no short sides
- no modern fade
- no mohawk
- no mullet

The main mass of hair must remain LONG, LOOSE, and FLOWING.

It should travel backward behind the ears and down across the shoulders.

Hair texture should feel:

- rugged
- windblown
- slightly damp
- naturally irregular
- individually stranded
- realistic
- heavy enough to have physical volume

Use the person's believable natural hair color family from the reference photograph, including gray where appropriate.

Do not make the hair unnaturally blond simply because this is a Viking portrait.


MANDATORY SIDE BRAIDS

Create TWO clearly visible narrow Viking braids framing the face:

- one narrow braid on the subject's left side
- one narrow braid on the subject's right side

Each braid begins near the temple or forward side of the scalp.

Each braid travels downward beside the face.

Each braid should extend approximately to the upper chest.

Both braids must remain visible.

Add one or more restrained engraved dark-metal Norse braid cuffs to each braid.

These braids are visually important.

Do NOT omit them.

Do NOT replace them with tiny barely visible braids.

Do NOT braid the entire hairstyle.

The majority of the hair remains loose.


REQUIRED VIKING BEARD - DO NOT OMIT

The final portrait MUST have a substantial Viking beard.

The beard must be significantly longer than normal contemporary facial hair.

The beard should extend well below the chin.

It should have:

- natural cheek growth
- realistic density variation
- individual hairs
- natural mustache integration
- rugged irregular texture
- realistic gray or lighter hairs where appropriate
- a longer central section beneath the chin

The beard must look as though it naturally grows from this person's actual facial anatomy.

Do not change the jaw or chin underneath it.


MANDATORY BEARD BRAIDS

The central lower beard must include THREE distinct narrow braided sections.

These beard braids are REQUIRED.

They should hang vertically beneath the chin.

Use:

- one central braid
- one slightly left of center
- one slightly right of center

Add small engraved Norse metal cuffs near the lower portions of the braids.

The braided portions should be clearly visible.

Do NOT replace the long braided beard with:

- short beard
- trimmed corporate beard
- goatee
- stubble
- generic full beard
- clean-shaven face

The side portions of the beard may remain naturally loose while the long center becomes braided.

The result should unmistakably read as a carefully groomed but rugged Viking chieftain beard.


FACIAL INTEGRATION

The face must look naturally integrated into the hair, beard, neck, fur, and environment.

The face should NOT look pasted onto another body.

Re-render the person's face completely under the environmental lighting.

Do not preserve the flat studio lighting from the reference photograph.

Hair must cast subtle natural shadows on the forehead and temples.

The beard must create realistic irregular shadows on the cheeks, jaw, and neck.

Hair around the ears must interact naturally with the ears.

Beard follicles must transition naturally into the beard.

Skin, hair, beard, fur, leather, and metal must share the same photographic sharpness, contrast, grain, and lighting environment.


EXPRESSION

Ignore a smile in the reference photograph.

Preserve facial identity, but change the expression.

Use:

- direct eye contact
- slightly narrowed eyes
- calm expression
- relaxed forehead
- closed mouth
- neutral lips
- controlled intensity
- subtle confidence

Do NOT show teeth.

Do NOT create a broad smile.

Do NOT create an angry grimace.

Do NOT scream.

Do NOT create theatrical aggression.

The expression should communicate:

experienced leader
intelligence
authority
calm confidence
controlled strength


SKIN

Render real human skin appropriate to an outdoor cold-weather portrait.

Include:

- pores
- beard follicles
- subtle pigmentation variation
- realistic fine lines
- age-appropriate texture
- minor natural imperfections
- subtle weather exposure
- restrained cold-weather redness

Preserve the person's apparent age.

Do not make the person significantly older.

Do not create leathery or cracked skin.

Do not exaggerate wrinkles.

Do not use beauty-filter skin.

Do not use waxy CGI skin.

Do not create excessive scars unless scars are visible in the reference image.


WARDROBE

Dress the person as a powerful, wealthy Viking chieftain.

Use dark, premium, historically inspired materials.

Include:

- dark brown leather
- nearly black leather
- layered leather armor
- subtle Scandinavian geometric tooling
- darkened metal rivets
- heavy leather straps
- dark textile layers
- subtle Norse engraving

Leather should show:

- real grain
- stitching
- age
- creasing
- subtle scratches
- handmade imperfections

The costume should feel like wardrobe from an extremely expensive prestige historical production.

It must NOT look like cosplay.


FUR MANTLE

Place a huge heavy fur mantle around both shoulders.

The fur should visually frame the face.

Use predominantly:

- very dark brown
- charcoal
- black-brown
- muted gray-brown

Keep the fur DARK.

Do not make it beige.

Do not make it golden brown.

Do not make it bright gray.

Do not use white fur.

Fur should appear:

- thick
- dense
- heavy
- slightly damp
- richly textured
- irregular
- individually stranded

The fur should occupy substantial portions of both sides of the portrait.


NORSE ORNAMENTATION

Include restrained high-status Viking ornamentation.

Use:

- large round engraved dark-metal brooches on the upper chest
- Scandinavian knotwork
- dark silver
- aged iron
- engraved braid cuffs
- dark beads
- leather cords

Metal should appear dark, aged, and hand-forged.

Avoid shiny modern metal.

Avoid excessive gold.

Avoid fantasy crowns.

Avoid horned helmets.


COMPOSITION - STRICT

Create a vertical 4:5 portrait.

This must be a TIGHT character portrait.

Frame approximately from mid-chest upward.

The face must be the dominant element.

The head should occupy a large percentage of the image.

The shoulders and fur should fill most of the lower frame.

Use approximately the visual perspective of a 100mm full-frame portrait lens.

Camera approximately eye level.

Subject almost straight toward camera.

Only slight shoulder rotation.

Do NOT zoom out to show the person's waist.

Do NOT create a full-body portrait.

Do NOT create an environmental portrait.

Do NOT introduce large foreground objects.

DO NOT INCLUDE:

- shield
- sword
- axe
- spear
- helmet
- bow
- large ship
- prominent longboat
- banners
- large fire
- large architecture

There should be NO shield anywhere in the image.

There should be NO weapon anywhere in the image.

The image is about:

FACE
HAIR
BEARD
FUR
LEATHER
ATMOSPHERE


BACKGROUND

Create a distant cold Scandinavian fjord environment.

The background must remain secondary.

It may contain only vague, heavily blurred suggestions of:

- dark mountains
- cold water
- mist
- distant shoreline
- indistinct settlement silhouettes
- two or three tiny orange fire lights

The background should be strongly out of focus.

Mountain detail should be subdued.

No object in the background should compete with the subject.

Do NOT create:

- bright snowfield
- bright village
- bright harbor
- prominent Viking ship
- readable banner
- large sail
- huge torch
- bright sky


LIGHTING - CRITICAL

The portrait must have DARK, LOW-KEY, CINEMATIC lighting.

This is one of the most important instructions.

The image should feel as though it was photographed during heavy Nordic overcast shortly before or after a storm.

Use:

- dark blue-gray ambient atmosphere
- soft directional light from camera-left/front-left
- restrained cool highlights
- deep but detailed shadows
- subtle cool environmental fill
- slight rim separation in selected hair strands
- subtle rim separation on fur

The face should emerge naturally from darkness.

The face should NOT be brightly illuminated like a studio portrait.

The face should NOT be dramatically brighter than the hair and fur.

Maintain meaningful shadow under:

- eyebrows
- eye sockets
- cheekbones
- nose
- mustache
- beard
- jaw

Eye catchlights should be very small and natural.

The shadow side of the face should remain visibly darker.

The overall scene should feel moody and underexposed compared with a normal commercial portrait.


TONAL TARGET - FOLLOW STRICTLY

Overall visual tone:

DARK
COOL
MOODY
DESATURATED
RICH
DRAMATIC
CINEMATIC

Dominant colors:

- deep slate blue
- dark blue-gray
- charcoal
- black-brown
- dark chocolate leather
- aged silver
- natural restrained skin tones

Background luminance should be significantly lower than a normal outdoor photograph.

Keep skies dark gray-blue.

Do NOT use a bright white cloudy sky.

Do NOT create bright snowy mountain peaks that dominate the frame.

Do NOT create warm sunny light.

Do NOT use golden-hour light.

Do NOT use orange-and-teal blockbuster grading.

Skin should remain naturally warmer than the surrounding environment but still clearly affected by the cool Nordic ambient light.

The final image should have deep blacks while retaining texture in fur and leather.


PHOTOGRAPHIC LOOK

The portrait should look like it was photographed with a premium cinema or medium-format camera.

Simulate approximately:

100mm portrait lens
f/2 to f/2.8
shallow optical depth of field

Keep:

- eyes extremely sharp
- eyelashes detailed
- facial hair detailed
- beard strands detailed
- front hair detailed
- gradual sharpness falloff
- strongly blurred distant mountains
- creamy natural bokeh
- realistic dynamic range
- physically plausible lighting

Avoid:

- HDR
- excessive clarity
- crunchy microcontrast
- oversharpening
- edge halos
- fake depth blur
- uniformly sharp background


CRITICAL FAILURE CONDITIONS

The image is WRONG if any of these occur:

- short modern hairstyle
- short beard
- missing beard braids
- missing side hair braids
- broad smile
- visible teeth
- bright face against darker body
- face looks pasted onto the body
- bright white sky
- bright snowy scenery
- large Viking ship
- shield
- weapon
- helmet
- wide environmental composition
- warm commercial portrait lighting
- beige or brightly colored fur
- generic Viking actor replacing the referenced person's identity


FINAL CHECK BEFORE GENERATING

Before creating the image, verify mentally that all of these are true:

The subject unmistakably resembles the reference person.

The hairstyle is LONG and swept backward.

Two clearly visible long side braids frame the face.

The beard is LONG.

Three visible braided beard sections hang beneath the chin.

Metal braid cuffs are visible.

The expression is calm and serious.

The fur is large, dark, and heavy.

The armor is dark brown and nearly black.

The portrait is tightly framed from approximately mid-chest upward.

There is no shield.

There are no weapons.

There is no prominent Viking ship.

The background is dark and heavily blurred.

The lighting is cool, dark, low-key, and cinematic.

The person's face looks naturally photographed within the environment rather than pasted into it.


FINAL CREATIVE TARGET

The result should feel like the definitive character portrait from a $150-million prestige Viking historical drama.

The viewer should see a dark, formidable Viking chieftain first.

Then, immediately upon looking at the face, someone who knows the person should think:

"Wow. That is absolutely them."

It should NOT feel like:

"a normal modern portrait with Viking clothing added."

It should feel like:

"this person actually existed in the Viking age and was photographed there."

Closing Thoughts

You can take almost any usable photograph, ask ChatGPT to help create a prompt from it, provide a reference image for the face, and then tweak the result to your heart's content. That is the real value here. It is a great starting point, and it does not take much to move from “interesting” to “okay, now we're getting somewhere.”

The caveat is simple: do not confuse the first impressive image with the finished image. If the mood is too clean, the beard is wrong, the texture is too smooth, or the personality is missing, keep working the prompt. In this case, I would add more grit to the second portrait in a follow-up comment. Give it a shot, slide the idea around, and see what happens.

This site uses cookies and affiliate links. If you’d like to support future posts, donate or use my coupon codes. All opinions are my own and do not represent any company.

No comments:

Post a Comment