February 2, 2026Tips & Tricks6 min read

Lighting Keywords That Transform Your AI Images

Master the vocabulary of light: golden hour, rim light, chiaroscuro, and more. Transform flat images into stunning photographs.

Which lighting keywords people actually use

Lighting advice usually arrives as a list of twenty terms with no indication of which ones anyone uses or what they do. This is a count instead.

We measured every lighting term across 9,599 published prompts in the LocalBanana gallery — each stored with the model that ran it and the image it produced. Below is the full table, in order, followed by what each term does to a face and which ones are sitting unused.

The full lighting keyword frequency table

Share of all 9,599 prompts containing each term, August 2026.

Lighting term	Share of prompts	What it does
studio lighting	6.5%	Controlled, even, shadowless by default. The safe answer
cinematic lighting	4.8%	High contrast plus colour separation between key and background
rim light	3.9%	A bright edge on the subject's outline; separates them from the background
natural light	3.8%	Whatever the scene would really have. Soft, directional, unbranded
flash	3.5%	Hard on-camera light, flattened faces, hot highlights, dark falloff
golden hour	2.7%	Low warm sun, long shadows, glowing edges
soft light	2.6%	Large source, gradual shadow edges, forgiving on skin
backlight	1.8%	Source behind the subject; halo hair, silhouette risk
warm light	1.5%	A colour temperature instruction, not a direction
overcast	1.5%	The sky as one enormous softbox. Flat, even, no drama
diffused light	1.3%	Same idea as soft, phrased as a modifier
volumetric light	1.3%	Visible beams; needs haze, steam or dust to exist
window light	1.1%	Soft and strongly directional at the same time
neon light	0.9%	Coloured practicals inside the frame
side lighting	0.5%	Light across the face; texture and shape
harsh light	0.3%	Small hard source, sharp shadow edges
moody lighting	0.3%	A mood word, not a setup — vague on purpose
blue hour	0.2%	After sunset; deep blue ambient against warm practicals
candlelight	0.2%	Small, warm, low, flickering; light from below
top light	0.05%	Directly overhead; eye sockets go dark

Two things are worth noticing before the details. First, the top of the table is the generic end: the two most-requested terms, studio lighting and cinematic lighting, are both closer to genre labels than to actual setups — neither one tells the model where the light is. Second, the bottom of the table is not there because those looks fail. top light at 0.05% is roughly five prompts in the entire corpus. Nobody thinks to ask.

The four terms that carry the corpus

studio lighting, 6.5%

The most-used lighting term is also the most default-shaped one. It gets you clean, even, commercially usable light — and it gets everyone else the same thing.

It earns its place when consistency matters — product shots, grids, character sheets. It is the wrong choice when you want the image to look like it happened somewhere.

cinematic lighting, 4.8%

A genre label rather than an instruction. It tends to produce contrast plus a colour split between key and background, which is why it looks good and why it looks the same everywhere. Naming the colours yourself gets you much further.

rim light, 3.9%

At 3.9% of the corpus, rim light is more common than natural light. Worth saying plainly: this is no longer an advanced technique, it is standard vocabulary. One prompt in twenty-five asks for it.

That has a consequence. If you add rim light expecting it to make your image distinctive, it will not — it will make it look like the several hundred other prompts that asked for the same edge. Use it when you actually need separation between subject and background, not as a general quality upgrade.

natural light, 3.8%

The most useful of the four, because it is the only one that does not import a production look. It says: light this the way the room would be lit.

flash at 3.5%: people are asking for a bad camera on purpose

flash appears in 3.5% of prompts, ahead of golden hour, soft light and every quality term below them. That is a genuine aesthetic shift showing up in the data. Direct on-camera flash — flat faces, hot foreheads, a hard shadow behind the shoulder, black falloff at the edges — was for decades the thing photographers spent money to avoid. It is now requested more often than soft light.

If you want it, ask for the mechanism rather than the mood:

Direct on-camera flash, small hard source, harsh specular highlights on
the forehead and cheekbones, hard-edged shadow on the wall behind the
shoulder, background falling two stops into darkness, slight
overexposure on the nearest surfaces, compact camera look

Time of day: golden hour at 2.7%, blue hour at 0.2%

golden hour is the most-used time-of-day instruction at 2.7%. It works — low warm sun, long shadows, glow on hair edges — and it is more than ten times as common as the hour that follows it.

blue hour sits at 0.2%. It is the twenty or so minutes after sunset when the ambient sky turns deep blue while every artificial light in the scene stays warm — the best free source of colour contrast in photography, and almost nobody asks for it.

Blue hour, twenty minutes after sunset, deep blue ambient sky as the
only fill, warm tungsten practicals inside the windows behind the
subject, no flash, colour contrast between cool skin shadows and warm
background lights, handheld, slight motion in the frame

Weather is a lighting instruction: overcast at 1.5%

overcast is a lighting setup disguised as a weather word — the entire sky becomes one enormous softbox. Even, directionless, no squinting, no hard shadow under the nose. It is why so much clothing catalogue work is shot on grey days.

At 1.5% it is badly under-asked relative to how much real photography depends on it. It is also the right answer whenever you want an image to read as unstaged — see lifestyle photography prompts.

Window light at 1.1%: soft and directional at once

Most soft light is directionless and most directional light is hard. A window is the common case where you get both, which is why it flatters faces at almost any angle. Specify the compass direction if you care — a north-facing window is even all day, a west-facing one turns orange in the evening.

The under-used end of the table

These four are each under 0.6% of the corpus. None of them are difficult; they are simply not in most people's vocabulary. Copy the lines.

side lighting, 0.5%

Light across the face rather than at it. Everything with texture becomes visible: skin, fabric, brick, hair.

Single source at 90 degrees to camera, light skimming across the face
from the left, right half falling into shadow with no fill, texture in
the skin and the knit clearly visible, hard shadow line down the centre
of the nose

harsh light, 0.3%

The look people spend money to avoid, used deliberately.

Through the Lens, the World Becomes Clear

Hard direct midday sunlight, no diffusion, no reflector, sharp-edged
black shadows, blown highlights on the shoulders, subject squinting
slightly, high contrast, colour saturated by the strong light

candlelight, 0.2%

Small, warm, low, and moving. The distinctive part is that it comes from below eye level, which inverts every shadow on the face.

Two candles on the table as the only light source, warm 1800K, light
falling on the underside of the chin and cheekbones, shadows inverted
upwards, deep falloff one metre from the flame, faint flicker, film
grain in the shadows

top light, 0.05%

The rarest term measured, roughly five prompts in the corpus. Overhead sources drop the eye sockets into darkness and put a bright line along the nose and the shoulders. Unflattering as a portrait default, extremely useful for anything that should look institutional, clinical or surveilled.

Overhead fluorescent strip directly above the subject, hard top light,
eye sockets in shadow, bright highlight along the nose bridge and the
tops of the shoulders, green colour cast, no fill, low-ceiling interior

Rare does not mean better

These terms are under-used, not secretly superior. Their value is that they produce an image which does not look like every other result for your subject — which matters exactly when your subject is a common one.

Two more terms worth knowing

neon light at 0.9% is the practical-light instruction: coloured sources visible inside the frame, lighting the subject from the scene rather than from a lamp you placed.

warm light at 1.5% is not a direction at all — it is a colour temperature. It pairs with any of the setups above rather than replacing them.

How to write a lighting line that survives

The terms above are shorthand. A lighting instruction that actually holds names four things, in this order:

Direction — where the source is relative to the camera. from camera left, 45 degrees behind the subject, directly overhead.
Quality — how large the source is, which decides how sharp the shadow edge is. large soft source, small hard source, bare bulb.
Ratio — what happens on the dark side. no fill, deep shadow, or white bounce filling the shadow to one stop under.
Colour — temperature or cast. warm 3200K, cool daylight, green fluorescent cast.

✕ Bad

beautiful dramatic cinematic lighting, moody atmosphere, professional

✓ Good

Single large softbox from camera left at 45 degrees and slightly above eye level, no fill on the right side, shadow falling to near black, warm 3200K key against a cool blue background wall

Four clauses, one per item on the list. Note that moody lighting — 0.3% of the corpus — is the term that survives the least well: it names an emotion and hands every actual decision back to the model.

Full example, ready to run:

Subject: man in his fifties, weathered face, grey stubble, plain dark
  work shirt, seated, forearms on the table
Setting: bare interior, plywood wall two metres behind him
Lighting: single small hard source from camera right at 60 degrees,
  slightly above eye level, no fill, shadow side falling to near black,
  warm 3000K key, one cool practical visible in the background
Camera: 85mm, f/2.8, eye level, chest-up, subject on the left third
Style: documentary portrait, natural skin texture, fine grain

Change only the Lighting block and rerun it in the generator. Swapping small hard source for large softbox half a metre from the face will change that image more than any other single edit available to you.

FAQ

What are the best lighting keywords for AI image generation?

For a face, the four that change the image most are side lighting, window light, backlight and harsh light — each one specifies a direction, which is what the model needs. studio lighting (6.5% of prompts) and cinematic lighting (4.8%) are the most-used, but they name a genre rather than a setup, so they produce the most generic results.

What does rim light do in an AI prompt?

It puts a bright edge along the subject's outline by placing a source behind them, which separates them from the background. It appears in 3.9% of prompts in our corpus — more often than natural light — so it is standard vocabulary rather than a distinguishing touch.

Why do so many AI images look over-lit or plastic?

Because studio lighting is the most requested term in the corpus at 6.5%, and studio light is by design even, shadowless and clean. Ask for a single source with no fill and let one side of the face go dark; that one change removes most of the plastic look.

Can I combine lighting keywords?

Yes, as long as they are not answering the same question twice or contradicting each other. golden hour plus backlight plus rim light is one coherent setup. golden hour plus overcast is two different skies, and the model will silently pick one. There is a full list of self-cancelling pairs in 10 prompt mistakes beginners make.

Where do these percentages come from?

They are exact counts over 9,599 published prompts in the LocalBanana gallery as of August 2026 — every item with a prompt longer than 20 characters. The wider analysis of the same corpus, including length and structure effects, is in what 9,667 AI prompts contain.

LocalBanana Team

We run Nano Banana, GPT Image and other models side by side, and publish every prompt in our gallery. These guides are written from that corpus.

Updated August 5, 2026@LocalBanana_io

Structured Prompt Templates: Six Slot Frameworks That Work

Tips & TricksAug 5, 20268 min read

Camera Language in AI Prompts: Lenses, Apertures, Composition

Tips & TricksAug 5, 20268 min read

How to Make AI Images Look Real, Not AI-Generated

Tips & TricksAug 5, 20268 min read

February 2, 2026Tips & Tricks6 min read

Lighting Keywords That Transform Your AI Images

Master the vocabulary of light: golden hour, rim light, chiaroscuro, and more. Transform flat images into stunning photographs.

Which lighting keywords people actually use

Lighting advice usually arrives as a list of twenty terms with no indication of which ones anyone uses or what they do. This is a count instead.

The full lighting keyword frequency table

Share of all 9,599 prompts containing each term, August 2026.

Lighting term	Share of prompts	What it does
studio lighting	6.5%	Controlled, even, shadowless by default. The safe answer
cinematic lighting	4.8%	High contrast plus colour separation between key and background
rim light	3.9%	A bright edge on the subject's outline; separates them from the background
natural light	3.8%	Whatever the scene would really have. Soft, directional, unbranded
flash	3.5%	Hard on-camera light, flattened faces, hot highlights, dark falloff
golden hour	2.7%	Low warm sun, long shadows, glowing edges
soft light	2.6%	Large source, gradual shadow edges, forgiving on skin
backlight	1.8%	Source behind the subject; halo hair, silhouette risk
warm light	1.5%	A colour temperature instruction, not a direction
overcast	1.5%	The sky as one enormous softbox. Flat, even, no drama
diffused light	1.3%	Same idea as soft, phrased as a modifier
volumetric light	1.3%	Visible beams; needs haze, steam or dust to exist
window light	1.1%	Soft and strongly directional at the same time
neon light	0.9%	Coloured practicals inside the frame
side lighting	0.5%	Light across the face; texture and shape
harsh light	0.3%	Small hard source, sharp shadow edges
moody lighting	0.3%	A mood word, not a setup — vague on purpose
blue hour	0.2%	After sunset; deep blue ambient against warm practicals
candlelight	0.2%	Small, warm, low, flickering; light from below
top light	0.05%	Directly overhead; eye sockets go dark

The four terms that carry the corpus

studio lighting, 6.5%

The most-used lighting term is also the most default-shaped one. It gets you clean, even, commercially usable light — and it gets everyone else the same thing.

It earns its place when consistency matters — product shots, grids, character sheets. It is the wrong choice when you want the image to look like it happened somewhere.

cinematic lighting, 4.8%

rim light, 3.9%

natural light, 3.8%

The most useful of the four, because it is the only one that does not import a production look. It says: light this the way the room would be lit.

flash at 3.5%: people are asking for a bad camera on purpose

If you want it, ask for the mechanism rather than the mood:

Direct on-camera flash, small hard source, harsh specular highlights on
the forehead and cheekbones, hard-edged shadow on the wall behind the
shoulder, background falling two stops into darkness, slight
overexposure on the nearest surfaces, compact camera look

Time of day: golden hour at 2.7%, blue hour at 0.2%

golden hour is the most-used time-of-day instruction at 2.7%. It works — low warm sun, long shadows, glow on hair edges — and it is more than ten times as common as the hour that follows it.

Blue hour, twenty minutes after sunset, deep blue ambient sky as the
only fill, warm tungsten practicals inside the windows behind the
subject, no flash, colour contrast between cool skin shadows and warm
background lights, handheld, slight motion in the frame

Weather is a lighting instruction: overcast at 1.5%

At 1.5% it is badly under-asked relative to how much real photography depends on it. It is also the right answer whenever you want an image to read as unstaged — see lifestyle photography prompts.

Window light at 1.1%: soft and directional at once

The under-used end of the table

These four are each under 0.6% of the corpus. None of them are difficult; they are simply not in most people's vocabulary. Copy the lines.

side lighting, 0.5%

Light across the face rather than at it. Everything with texture becomes visible: skin, fabric, brick, hair.

Single source at 90 degrees to camera, light skimming across the face
from the left, right half falling into shadow with no fill, texture in
the skin and the knit clearly visible, hard shadow line down the centre
of the nose

harsh light, 0.3%

The look people spend money to avoid, used deliberately.

Hard direct midday sunlight, no diffusion, no reflector, sharp-edged
black shadows, blown highlights on the shoulders, subject squinting
slightly, high contrast, colour saturated by the strong light

candlelight, 0.2%

Small, warm, low, and moving. The distinctive part is that it comes from below eye level, which inverts every shadow on the face.

Two candles on the table as the only light source, warm 1800K, light
falling on the underside of the chin and cheekbones, shadows inverted
upwards, deep falloff one metre from the flame, faint flicker, film
grain in the shadows

top light, 0.05%

Overhead fluorescent strip directly above the subject, hard top light,
eye sockets in shadow, bright highlight along the nose bridge and the
tops of the shoulders, green colour cast, no fill, low-ceiling interior

Rare does not mean better

Two more terms worth knowing

neon light at 0.9% is the practical-light instruction: coloured sources visible inside the frame, lighting the subject from the scene rather than from a lamp you placed.

warm light at 1.5% is not a direction at all — it is a colour temperature. It pairs with any of the setups above rather than replacing them.

How to write a lighting line that survives

The terms above are shorthand. A lighting instruction that actually holds names four things, in this order:

Direction — where the source is relative to the camera. from camera left, 45 degrees behind the subject, directly overhead.
Quality — how large the source is, which decides how sharp the shadow edge is. large soft source, small hard source, bare bulb.
Ratio — what happens on the dark side. no fill, deep shadow, or white bounce filling the shadow to one stop under.
Colour — temperature or cast. warm 3200K, cool daylight, green fluorescent cast.

✕ Bad

beautiful dramatic cinematic lighting, moody atmosphere, professional

✓ Good

Single large softbox from camera left at 45 degrees and slightly above eye level, no fill on the right side, shadow falling to near black, warm 3200K key against a cool blue background wall

Full example, ready to run:

Subject: man in his fifties, weathered face, grey stubble, plain dark
  work shirt, seated, forearms on the table
Setting: bare interior, plywood wall two metres behind him
Lighting: single small hard source from camera right at 60 degrees,
  slightly above eye level, no fill, shadow side falling to near black,
  warm 3000K key, one cool practical visible in the background
Camera: 85mm, f/2.8, eye level, chest-up, subject on the left third
Style: documentary portrait, natural skin texture, fine grain

FAQ

What are the best lighting keywords for AI image generation?

What does rim light do in an AI prompt?

Why do so many AI images look over-lit or plastic?

Can I combine lighting keywords?

Where do these percentages come from?

LocalBanana Team

We run Nano Banana, GPT Image and other models side by side, and publish every prompt in our gallery. These guides are written from that corpus.

Updated August 5, 2026@LocalBanana_io

Which lighting keywords people actually use

The full lighting keyword frequency table

The four terms that carry the corpus

studio lighting, 6.5%

cinematic lighting, 4.8%

rim light, 3.9%

natural light, 3.8%

flash at 3.5%: people are asking for a bad camera on purpose

Time of day: golden hour at 2.7%, blue hour at 0.2%

Weather is a lighting instruction: overcast at 1.5%

Window light at 1.1%: soft and directional at once

The under-used end of the table

side lighting, 0.5%

harsh light, 0.3%

candlelight, 0.2%

top light, 0.05%

Two more terms worth knowing

How to write a lighting line that survives

FAQ

What are the best lighting keywords for AI image generation?

What does rim light do in an AI prompt?

Why do so many AI images look over-lit or plastic?

Can I combine lighting keywords?

Where do these percentages come from?

Related articles

Structured Prompt Templates: Six Slot Frameworks That Work

Camera Language in AI Prompts: Lenses, Apertures, Composition

How to Make AI Images Look Real, Not AI-Generated

Which lighting keywords people actually use

The full lighting keyword frequency table

The four terms that carry the corpus

studio lighting, 6.5%

cinematic lighting, 4.8%

rim light, 3.9%

natural light, 3.8%

flash at 3.5%: people are asking for a bad camera on purpose

Time of day: golden hour at 2.7%, blue hour at 0.2%

Weather is a lighting instruction: overcast at 1.5%

Window light at 1.1%: soft and directional at once

The under-used end of the table

side lighting, 0.5%

harsh light, 0.3%

candlelight, 0.2%

top light, 0.05%

Two more terms worth knowing

How to write a lighting line that survives

FAQ

What are the best lighting keywords for AI image generation?

What does rim light do in an AI prompt?

Why do so many AI images look over-lit or plastic?

Can I combine lighting keywords?

Where do these percentages come from?

Related articles

Structured Prompt Templates: Six Slot Frameworks That Work

Camera Language in AI Prompts: Lenses, Apertures, Composition

How to Make AI Images Look Real, Not AI-Generated