Three Commands, One Picture: How /you, /pic and /image Differ

You can ask a companion for a picture, and there are three ways to do it. They look interchangeable and they are not: they differ on exactly one axis, decided before anything creative happens. Knowing which is which saves you the round trip of getting a lovely image of a place with nobody in it. Here is what each command does, what the allowance is, and where the line sits on what can be made.

The One Axis

All three commands take a description and produce an image. The only difference between them is whether your companion is in the picture.

So /you at sunset gives you your companion at sunset. /image at sunset gives you a sunset. Both are correct behaviour and the difference is entirely in which command you reached for.

The subjectless phrasing of /you is the reason it exists. Describing a scene without having to build a sentence around the person in it composes more naturally than the alternatives, and it turned out to be what people typed anyway.

How /image Decides

This is the part worth knowing precisely, because a guess here would be an annoying thing to guess about.

/image includes your companion when your description refers to them: a word like you, your, yourself, us or we, or their name. That check is a plain word match, it runs on the server before anything else happens, and it is deterministic. The decision is never left to a model's reading of what you probably meant.

The practical upshot: /image us on a beach includes them. /image a beach does not. If you want certainty rather than a rule to remember, use /you, which never asks the question.

What It Costs

Every tier can use all three commands. The difference between tiers is how many pictures you get.

The free allowance is worth a sentence because the shape of it is unusual. It is per source rather than shared: the dashboard portrait and the in-chat picture draw from separate pools of three. Using up your chat pictures does not touch your ability to make a portrait, and the reverse holds too.

That number moved upward recently, which is the opposite of the usual direction, and the reasoning is worth stating. A first attempt at a picture frequently misses: the description is read differently than you pictured it, or it is simply not what was in your head. At one per source, a single near-miss meant a person's entire experience of the feature was a disappointment with no way to try again. Three is enough to learn what the description does, adjust, and land on something. It is still far too few to live on, which is honest rather than accidental.

A Refusal Does Not Cost You a Picture

Sometimes a request is declined. When that happens on the free tier, it no longer consumes your allowance.

That was not always true, and the old behaviour was hard to defend once it was written down plainly: a person was being charged for the product saying no to them, and if it happened on a first attempt the result was no pictures and nothing left to try with. The grace is bounded rather than unlimited, covering up to your allowance rather than indefinitely, so a description that keeps getting declined cannot be repeated forever at no cost. That bound is deliberate and it is stated here rather than discovered.

Where the Line Is

Pictures are kept to what the product describes as safe for work, and the way that line is drawn changed in a way worth explaining, because the change made it work better rather than merely differently.

The gate blocks what is being asked for, not the words used to ask. Nudity, explicit acts and requests for exposure are declined. Ordinary descriptive language about a body is not, on its own, a reason to decline, and requests using it are rendered clothed. The earlier behaviour turned on vocabulary, which produced the worst of both outcomes: it declined perfectly ordinary requests while being trivially routed around by anybody phrasing things differently.

One property of that change is worth stating explicitly. Loosening the picture gate did not loosen anything about protecting minors. Those checks are defined so that words leaving the picture gate cannot leave that one, and the relationship between them is pinned by tests so it cannot drift apart later. That is a structural guarantee rather than a policy promise.

How Your Companion Stays Themselves

One last mechanism, because it answers the question people ask after their second picture: why does the same companion look like the same person across different scenes?

Appearance comes from what is saved about them (the structured description built when they were created, plus whatever you have written in the appearance fields) and from a reference to their existing image, so each new picture is anchored to the last rather than being invented afresh.

Something that used to feed into this and no longer does: the companion's written backstory. It was being cut to a fixed length and passed along, which meant a sentence chopped off mid-word, describing a relationship rather than a face, was arriving as though it were a visual instruction. It contributed nothing to how anyone looked and occasionally caused a request to be declined for reasons that had nothing to do with the picture. Removing it made results both better and more predictable. Character description belongs to the conversation; appearance belongs to the appearance fields.