Skip to content
Generative portrait with an alternative aesthetic in violet and cyan, backlit through haze.

Proyecto propio · @littlefenu en Threads2025–2026Case study

Generative series for Threads

Generative image series published on Threads: one concept, four variants and a prompt written like a camera spec sheet.

Client
Proyecto propio · @littlefenu en Threads
Year
2025–2026
Role
Creative direction and AI generation
Categories
Generative AI · Editorial
Tools
Gemini (Nano Banana) · Midjourney v6 · Stable Diffusion XL · DALL·E 3 · Photoshop · ChatGPT

The project

The brief

Threads is where I publish the generative work in the open. Each series starts from a concept — a light, a texture, a mood — and resolves into three or four images published together, with the prompt visible. It is not an idle experiment: it is the lab where I test what later goes into brand commissions.

The method never changes, and it is what separates a usable result from a model-gallery curiosity. I write the prompt the way I would write a spec sheet: focal length, aperture, ISO, lighting scheme, reference film stock, contrast and aspect ratio. If the image fails I do not switch models — I fix the parameter that is wrong, the same way I would fix an exposure.

I generate in Gemini (Nano Banana), Midjourney, Stable Diffusion XL and DALL·E 3 depending on what the piece needs — Midjourney for character, SDXL when I need fine control, Nano Banana for editing on top of a real photograph — and I always finish in Photoshop: a model output is a negative, not a delivery.

My contribution

  • I define the concept for each series and lock it in a mood board before writing a single line of prompt.
  • I write prompts with real photographic parameters — focal length, aperture, ISO, lighting scheme, grain — instead of adjectives.
  • I run the same scene across models and pick whichever solves that specific piece, not the habitual one.
  • I iterate one parameter at a time, so I know what each change produces and can repeat it.
  • I finish every output in Photoshop: colour, density, grain and artefact cleanup.
  • I publish each series with the prompt visible and the model disclosed.
Base. The same frame in two states: the photographic base on the left, the generative output on the right.

Base

AI. The same frame in two states: the photographic base on the left, the generative output on the right.

AIMidjourney v6

The base sets the framing, the focal length and the direction of the light; the model only resolves the atmosphere.

Square piece from the series: figure cut out against a violet halo.
Square piece from the series: cyan light geometry on a deep background.
Square piece from the series: profile portrait with a hard cut-out.

The model knows no photography. If you do, the prompt stops being a description and becomes a call sheet.

Patricia Fenutria · nota de proceso

Impact

The results

  • 4Models in rotation, chosen per piece
  • 3–4Images per published series
Back to work
Beauty close-up lit with an octabox, skin texture left intact.

Up next

On to Beauty and product: still life, texture and skin

Next project

Contact

Let's talk.

contact@patriciafenutria.com