This Is What Nomi AI’s Image Generator Produces

In just the past few days, two different posts on the official Nomi.ai subreddit showcased images generated directly by the platform’s…

This Is What Nomi AI’s Image Generator Produces

In just the past few days, two different posts on the official Nomi.ai subreddit showcased images generated directly by the platform’s internal image engine.

In the first post, a gallery of eight generated images included at least two figures who appeared visibly underage. In a separate post, another generated figure was depicted in a highly sensual pose on a bed.

(Note: The original version of the third image depicted the figure in underwear. We have manually applied censorship bars to cover the exposed skin before sharing it here.)

To remove subjective bias from the observation that these avatars look like adolescents, the images were run through an external facial age-analysis tool. The objective data confirms the visual impression.

As the external analysis indicates, the figures generated by the platform are assessed as highly likely to be minors (in the 15–18 age range).

This is not a debate about user prompts. This is a direct look at what the platform’s image generator is permitted to create and routinely outputs. The system actively generates avatars with adolescent facial features and proportions, even when placing them in suggestive or sensual contexts.

It is also worth noting how the platform attempts to manage this. Users frequently report innocent image prompts being mysteriously rejected. The community usually discovers that this is due to a crude, hidden blacklist of specific phrases — such as blocking the words “is one” or “young girl.”

Why rely on such strict word bans? Because the developers know exactly what their underlying image model is truly capable of producing if left unchecked.

However, this filter is purely performative. As the images above prove — and as users have explicitly complained when platform updates suddenly “de-aged” their adult companions to look 16 — the system still actively outputs minor-coded avatars, often entirely unprompted. Banning a few specific words does not fix a generative engine that is structurally biased and allowed to produce “youth”.