[Experiment] Lens Focal Length Prompt Comparison | How Does Output Change from 24mm to 135mm?

Conclusions First

• Fisheye lens (fisheye) clearly reproduced barrel distortion and ultra-wide angle in all 3 out of 3 — the most impactful focal length specification.
• 24mm wide angle showed a tendency for slightly wider framing and more full-body composition, but perspective distortion reproduction was unstable.
• In the 35mm to 135mm standard to telephoto range, no consistent difference in background bokeh, compression effect, or framing was confirmed compared to control. Subject framing and background depiction fall within seed-to-seed variation.
• With the 85mm f/1.4 “shallow depth of field, creamy bokeh” specification, background bokeh increased slightly in only 1 out of 3 — f-value specification for depth of field control is unstable.

Experiment Design

Base Prompt

A cafe interior with depth was used as the background to make focal length-based background compression and bokeh changes easier to observe. Upper body framing also allows observation of perspective changes in the subject.

Prior Knowledge

From prompt best practices, the following has been confirmed:

• Camera model names (Canon EOS R5, etc.) have no effect
• This article tests whether “focal length numerical specification” affects composition and bokeh, rather than model names

Evaluation Criteria

The following image attributes were observed and recorded:

• Subject perspective: Face distortion, proportion changes (wide-angle peripheral stretching, etc.)
• Background bokeh/compression: Size, detail, and blur degree of background objects
• Field of view width: Amount of environmental depiction in the surroundings

Results for Each Condition

A00: Control (no lens specification)

seed 1	seed 2	seed 3

Observation: All 3 show an upper body composition in front of the cafe. The background depicts the cafe interior (shelves, cups, coffee equipment) through glass with moderate detail. No face distortion with standard perspective. Background bokeh is light with most objects still identifiable.

A01: 24mm Wide Angle

seed 1	seed 2	seed 3

Observation: All 3 out of 3 show wider framing than control, with the full body or below the waist visible. More environmental depiction in the background including cafe exterior, terrace seating, and trees. However, wide-angle characteristic peripheral stretching or face distortion is not confirmed — the perspective itself is not substantially different from control.

A02: 35mm f/1.4

seed 1	seed 2	seed 3

Observation: No clear difference from control. All 3 show upper body framing with the same cafe background detail and bokeh as control. No change in subject perspective is observed. The bokeh enhancement suggesting f/1.4’s shallow depth of field was also not confirmed.

A03: 50mm f/1.4

seed 1	seed 2	seed 3

Observation: No clear difference from control. Framing, background depiction, and bokeh are all within control range. All 3 show standard upper body composition, and the 50mm’s “natural perspective” is indistinguishable from control.

A04: 85mm f/1.4

seed 1	seed 2	seed 3

Observation: Seed 1 shows slightly stronger background bokeh than other conditions, with the glass-view interior details softened. However, seeds 2 and 3 show the same background depiction as control. Framing is the same upper body as control. Background compression (telephoto effect making background objects appear larger) is not confirmed in any of the 3. The influence of the “shallow depth of field, creamy bokeh” added text is limited to 1 out of 3.

A05: 135mm f/2

seed 1	seed 2	seed 3

Observation: Seed 1 depicts 2 people despite the “1girl” specification — broken composition. Seeds 2 and 3 show upper body composition with cafe interior background. Background compression and bokeh enhancement are not confirmed compared to control. Telephoto lens’s narrow angle of view (subject cropped large) is also not seen, falling within seed-to-seed variation.

A06: Fisheye Lens

seed 1	seed 2	seed 3

Observation: All 3 out of 3 clearly reproduce barrel distortion. The image periphery curves, with straight lines (ceiling, walls, counter) bowing outward. The field of view is extremely wide with ceiling, floor, and both walls of the cafe interior fitting in a single frame. Subject is centered with the head appearing relatively large and feet small — an overhead perspective in 2 out of 3. Seed 2 shows a black circular vignette, mimicking the optical characteristics of a fisheye lens. All 3 subject faces are rendered without distortion.

Note: Reproducing not just the barrel distortion but also the vignette is a notable result. It is possible, though unconfirmed, that the training data includes a substantial amount of fisheye lens imagery.

Summary

Recommended Prompts

The only focal length specification that produces stable effects is fisheye only. Focal length specifications in the standard to telephoto range (35mm-135mm) show no difference from control and are likely to be a waste of tokens.

The 24mm wide angle shows a partial effect of widening the field of view, but since the wide-angle perspective distortion isn’t reproduced, if you just want a pulled-back composition, full body or other framing specifications are more reliable.

For stronger background bokeh, direct visual effect words like shallow depth of field or bokeh are likely more effective than focal length or f-value specifications (not tested in this experiment).

Related Articles

[Verified] Image Generation Prompt Best Practices

[Experiment] 6 Camera Angle Comparison — How Does a Poolside Scene Change with Low Angle, Bird's Eye, and Dutch Angle?