Companion Selfies, and Listening to Our Early Users
30 May 2026We're coming close to the end of the month now, and we've been hard at work figuring out what features we must prioritize, what our ideal user flows should look like, and how to reach out to more early users.
Most of all we want to give a shoutout to our Day Ones - our early users who spent hours texting on our unpolished app. We spent hours speaking to early users to understand more about your daily lives, your world views, and your past experiences with AI companions. We received 10-page Word documents with feature requests, bug reports, all fully documented with screenshots and arrows. We had hour-long text conversations with many of you over Discord, Reddit, and Telegram. And most of all, we're excited to do right by our users and build our ProjectSylvie to be the most realistic text messaging experience.
Companion selfies¶
First off, we've just pushed a new feature allowing users to request their selfies from their Companions. Tap the camera button in any chat and describe the scene you want to see "you at the beach at golden hour," "cozy in a sweater with a coffee," "dressed up for a night out." A few seconds later, a photo of the user's Companion lands in the conversation, using their Companions' profile picture as a reference pic. Companions don't yet understand the context of the pictures they're sending, but that's a work in progress.
While the photo is generating, a loading bubble holds its place in the conversation, and it survives page reloads, so you can wander off and come back to find it waiting.
Custom backgrounds¶
While we were making chats more visual, we added a function to allow users to set a custom background for each Companion conversation window. Upload an image from your Companion's profile page, frame it how you like, and it becomes the backdrop of your chat.
Image generation tokens¶
Selfie generation uses a few message credits per photo, with each plan's message balance covering everything in one place. We built it this way to keep the credits system simple and intuitive for the user. Failed generations are not charged, and generated images are stored privately to each user.
The camera function must point in both directions, and we intend to release both user image uploads and Companion recognition of user images (as well as Companion selfies) very soon, so do watch out for our next update!