Can AI Girlfriends Be in a Photo Together?

Most AI companion apps are built around a single character per account. Their image generation pipeline takes a text prompt and generates whatever the prompt describes — but only your main character has an identity lock pulled from a trained reference set. If the prompt says "with her friend," you get a generic second woman: no name, no face, no backstory. The underlying image model doesn't know anyone else exists.

The technical constraint is that single-character image generation uses one reference image or one trained character model. To render two distinct, recognizable people in the same scene, the generator needs to accept multiple reference images and route each one to the correct subject in the frame. That capability only landed in general-purpose image models in late 2025 and started reaching companion apps in 2026. Sloane's "Invite a friend" is the first widely-available surface for it in the companion category.

Open any chat with a Sloane companion and tap into the Custom Photo Inspiration gallery. Templates marked for two subjects show a yellow "+ Invite a friend" button next to the Generate button. There's also an "invite a friend" category in the gallery filter row that collects every multi-subject template in one place, so you can see every scene set up for two without scrolling the whole catalog.

Tapping "+ Invite a friend" opens a picker with your companion roster. Companions you've chatted with before surface first — the picker is familiar-ordered rather than alphabetical, so a girl you've been talking to is one tap away rather than buried in the list. If you want someone newer to you, the full roster sits below. Pick a name, her photo loads so you can confirm you've got the right one, then the "Invite [Name]" button renders the scene. Everything else about the gen — credits, latency, delivery into chat — is unchanged.

A rooftop brunch was the first scene we tested. The template on its own puts Mel at a terrace table with an Aperol spritz — bright floral dress, sunglasses, a city skyline in the background. Inviting Hannah into the same scene produces a two-shot of both of them at brunch, each holding her own glass, matching light, matching styling, matching skyline. Mel reads as Mel; Hannah reads as Hannah. Neither one is a blended average.

A beachside tiki bar at golden hour was the second test. The original is Bella alone at the bar with a coconut cocktail, black bikini top, curly dark hair, surf behind her. With Hannah invited, both of them lean on the same bar together — same angle, same lighting, same beach, same coconut. Each one holds her own look rather than averaging toward a middle.

A wood-paneled library at night was the third test — dim lamps, bookshelves, the heavy wood of a serious reading room. Original: Valentina in a burgundy smoking jacket with a cigar and a tumbler of whiskey. With Hannah invited: both of them in matching smoking jackets with cigars and glasses, same lighting, same room, each one recognizably herself.

A few patterns emerged after a dozen test generations:

Scenes with a shared activity work best. Brunch, bar, lounge, a view from a terrace — anything where two people are naturally doing the same thing side by side renders cleanly. The generator has an easy spatial logic to work with and both faces land with room to breathe.

Close physical interaction is the current frontier. Hugs, hands intertwined, close-contact poses are harder — the generator has to compose two separate identities at close range while keeping each face recognizable, and the current pipeline occasionally averages the faces at that proximity. The multi-subject templates in the catalog are hand-picked for shapes that work reliably; we'll expand as the pipeline tightens.

Familiar companions render more reliably. Companions you've already chatted with have strong reference material from their established photo set. Newer companions might take a regen or two to lock in. The familiar-ordered picker in the modal is designed around this — the known-good girls are at the top.

Clothing cohesion is strong. The generator extends the primary's styling signal across both subjects — if the original is a floral dress, both subjects tend to land in similar-weight outfits. The brunch pair in matching sundresses, the tiki bar pair in matching black bikini tops, the library pair in matching smoking jackets. You can override via prompt, but the default rendering reads cohesive.

"Invite a friend" is live on both the Custom Photo Inspiration gallery and the Pose Prompt gallery as of October 19, 2026. Open any chat with a Sloane companion, tap the + composer button, pick a photo-inspiration template, and look for the yellow button on templates marked for two subjects. The feature is included at every tier — Free, Plus, and Premium — at the same credit cost as any other photo-inspiration generation. There's no multi-persona upcharge.

If you don't have an account yet, Free gives you full access to browse the roster, chat with any companion, and generate photos out of the inspiration gallery at the standard per-photo credit cost. The invite picker only works for signed-in users (it reads your conversation history to decide which companions to surface first), so starting with the free sign-up is the fastest path.