When drafting prompts for image generation models, many creators routinely append technical hardware specs to their sentences: "Shot on Sony A7R V," "35mm f/1.4 lens," or "f/11 aperture."
A persistent question among designers and prompt engineers is whether modern text encoders actually interpret real-world optical physics. Do specific focal lengths and f-stops trigger genuine lens blur and perspective compression, or do they function as aesthetic placebos with no measurable effect?
To test the practical impact of camera parameters, we ran two controlled benchmark experiments. We kept environmental descriptions and framing identical while varying only the camera body, focal length, aperture, shutter speed, and ISO.
Experiment 1: New York architectural intersection
For our first test, we selected a complex urban cityscape featuring historic cast-iron facades contrasted against modern glass skyscrapers. We held the composition, street-level perspective, and lighting fixed, changing only the photographic hardware values across four variations.
Test prompts
Prompt 1: Sony ultra-wide 16mm
An architectural photograph of a New York Manhattan intersection at dusk, featuring historic cast-iron brick facades contrasted against sleek glass modern skyscrapers, rain-slicked asphalt reflecting traffic lights and street signs, thin white steam rising from a street manhole cover, classic iron fire escapes on building exteriors, eye-level street perspective, straight-on architectural composition, no people in the scene. Shot on Sony A7R V, 16mm f/2.8 lens, f/11 aperture, ISO 100, 1/60s shutter speed.Prompt 2: Leica standard 35mm
An architectural photograph of a New York Manhattan intersection at dusk, featuring historic cast-iron brick facades contrasted against sleek glass modern skyscrapers, rain-slicked asphalt reflecting traffic lights and street signs, thin white steam rising from a street manhole cover, classic iron fire escapes on building exteriors, eye-level street perspective, straight-on architectural composition, no people in the scene. Shot on Leica M11, 35mm f/1.4 Summilux lens, f/4 aperture, ISO 200, 1/125s shutter speed.Prompt 3: Nikon telephoto 200mm
An architectural photograph of a New York Manhattan intersection at dusk, featuring historic cast-iron brick facades contrasted against sleek glass modern skyscrapers, rain-slicked asphalt reflecting traffic lights and street signs, thin white steam rising from a street manhole cover, classic iron fire escapes on building exteriors, eye-level street perspective, straight-on architectural composition, no people in the scene. Shot on Nikon Z9, 200mm f/2.8 VR lens, f/2.8 aperture, ISO 400, 1/250s shutter speed.Prompt 4: Hasselblad medium format 80mm
An architectural photograph of a New York Manhattan intersection at dusk, featuring historic cast-iron brick facades contrasted against sleek glass modern skyscrapers, rain-slicked asphalt reflecting traffic lights and street signs, thin white steam rising from a street manhole cover, classic iron fire escapes on building exteriors, eye-level street perspective, straight-on architectural composition, no people in the scene. Shot on Hasselblad H6D-100c, 80mm f/2.8 lens, f/8 aperture, ISO 64, 1/30s shutter speed.Experiment 1 results: No observable optical shift
Comparing the four generated cityscapes side by side revealed no meaningful visual divergence across the set.
Despite explicitly requesting a 16mm ultra-wide lens, the field of view did not widen, nor did vertical lines show barrel distortion. Similarly, specifying a 200mm telephoto lens failed to compress background buildings or create telephoto cropping.
Experiment 2: Tabletop still life in an antique study
Our cityscape results suggested that wide scenes with uniform spatial depth make focal shifts difficult for models to map.
To test this hypothesis, we set up a second experiment around a tabletop macro subject with distinct depth layers: a vintage brass pocket watch on an aged oak table, with stacked leather books and a glowing desk lamp in the background.
Test prompts
Prompt 1: 100mm macro lens with wide f/2.8 aperture
A still life photograph of a vintage mechanical brass pocket watch resting on an aged dark oak wooden table in an antique library, with stacked leather-bound books and a warm glowing desk lamp in the background, eye-level tabletop perspective, centered composition, no people in the scene. Shot on Canon EOS R5, 100mm f/2.8 Macro lens, f/2.8 aperture, ISO 100, 1/125s shutter speed.Prompt 2: 16mm ultra-wide lens with narrow f/16 aperture
A still life photograph of a vintage mechanical brass pocket watch resting on an aged dark oak wooden table in an antique library, with stacked leather-bound books and a warm glowing desk lamp in the background, eye-level tabletop perspective, centered composition, no people in the scene. Shot on Sony A7R V, 16mm f/2.8 lens, f/16 aperture, ISO 100, 1/30s shutter speed.Prompt 3: 35mm analog film camera with Portra 400 film stock
A still life photograph of a vintage mechanical brass pocket watch resting on an aged dark oak wooden table in an antique library, with stacked leather-bound books and a warm glowing desk lamp in the background, eye-level tabletop perspective, centered composition, no people in the scene. Shot on Leica M6 35mm film camera, 35mm f/1.4 lens, f/4 aperture, Kodak Portra 400 film stock.Experiment 2 results: Clear optical separation
In the tabletop still life setup, the outputs diverged significantly. The three prompts produced distinct variations in depth of field, background roll-off, and highlight transitions.
Prompt 1 rendered a shallow depth of field. Focus locked sharply onto the pocket watch face and oak wood grain, while the desk lamp and bookshelves dissolved into smooth background blur.
Prompt 2 accurately reflected the narrow f/16 aperture by applying deep pan-focus. Every object from the foreground dial to the background books remained crisp. However, removing optical blur gave the overall render a flatter, slightly illustrative look.
Prompt 3 sat between the first two outputs. The background remained readable but secondary. Highlight transitions across the polished brass watch softened, yielding a natural analog film texture.
Why camera prompts behave differently across scene types
Why did camera specifications produce zero optical change in the cityscape, yet create distinct physical differences in the still life setup?
In complex landscape or architectural environments, the high density of environmental tokens (skyscrapers, asphalt, steam, fire escapes, traffic lights) dominates the text encoder's attention weights. Technical camera values placed at the end of the prompt become secondary to the dominant structural descriptions.
In contrast, a tabletop still life provides an explicit spatial hierarchy between a hero foreground subject and a distant background. This distinct depth structure gives the diffusion pipeline an actionable spatial plane onto which it can map focal length and aperture instructions.
Practical workflow takeaways
Our benchmark tests point to three practical rules for working with camera parameters:
Use camera and aperture tags for close-up and portrait work: When your scene features a clear foreground subject, specifying a "100mm macro lens" or "f/1.8 aperture" reliably controls background bokeh and subject separation.
Do not rely on lens tags to force wide perspective in landscapes: Specifying a "16mm wide-angle lens" will not force a landscape scene to widen if the text describes a full street layout. To change framing in wide scenes, use explicit compositional terms like "extreme wide-angle establishing shot" or "distant aerial perspective."
Use camera brand tags as aesthetic quality anchors: While writing "Leica M11" or "Hasselblad H6D" might not replicate specific physical glass physics, it anchors the prompt within high-end commercial photography latent clusters, steering the generator away from amateur snapshot aesthetics.
If you are prototyping complex commercial mockups, save these prompt structures to your ctrlc.ai collection to quickly test different focal setups across your projects.
