A bit more dynamic and less human focused comparison between GPT 1.5 and nano banana pro

Understanding the Challenges of Precision in AI-Generated Photography Descriptions: An Analytical Comparison of GPT 1.5 and Nano Banana Pro

In the rapidly evolving landscape of artificial intelligence and text-to-image generation, clarity and control over prompt inputs remain pivotal for achieving desired visual outcomes. This article examines a comparative analysis between two AI models—GPT 1.5 and Nano Banana Pro—highlighting their responses when tasked with generating detailed photographic prompts, specifically focusing on adherence to technical parameters such as focal length.

The Scenario

The challenge involves instructing the AI to produce a photographic description that accurately reflects the use of a 24mm focal length lens, a common specification for capturing wide-angle images with an expansive field of view. The prompt provided encapsulates specific stylistic and technical elements, aiming to produce a raw, authentic, and understated documentary-style photo.

Sample Prompt

“Raw ultra-realistic candid amateur photo taken on an Android smartphone, 24mm wide lens. Natural lighting, no dramatic shadows. Low contrast, slightly washed-out muted colors. Everyday mundane moment, boring real life vibe. Everything in focus with deep depth of field, no bokeh. Casual, slightly off-center framing. Unedited and unpolished with minor flaws like mild noise/grain, uneven exposure, subtle motion softness, and slight JPEG compression artifacts. Authentic early-2020s phone snapshot realism.”

Analysis of AI Responses

Despite clear instructions emphasizing the 24mm focal length, the models exhibited notable limitations:

• GPT 1.5: The model showed a tendency to overlook the specific focal length command, resulting in descriptions that, although detailed, lacked consistent emphasis on the wide-angle perspective. Its generation often defaulted to generic smartphone photography descriptions without explicitly linking elements like depth of field and focal length parameters.

• Nano Banana Pro: Demonstrated a somewhat better grasp of technical details but still struggled to incorporate the requested focal length into the description seamlessly. The model tended to focus more on stylistic cues rather than technical accuracy, occasionally producing imagery that did not fully align with the stipulation of a 24mm lens.

Key Observations

1. Focal Length Specification: Neither model reliably integrated the 24mm focal length parameter into their descriptive outputs, highlighting an area where AI interpretive capabilities can be improved.

2. Stylistic Fidelity: Both models successfully captured the intended aesthetics—authenticity, simplicity, and understated realism—though with varying degrees of technical adherence.