A letter from the maker.

I've always been a vivid dreamer. For years, I'd wake with something luminous still behind my eyes. A conversation more real than most I have awake. A place I was sure I'd been before. By the time I reached for my phone, though, it had all evaporated. Not blurred. Gone. I tried keeping a notebook by the bed. I tried voice memos. Neither worked. What I needed was something that would meet me in that half-asleep moment, gently pulling the dream out through conversation, before the waking brain could edit it away. So I made it.

Dreams arrive as images first. Words come later. The moment we put language to a dream, we're already translating, and some of the dream gets lost in the translation. The visualization step in Epona is an attempt to shorten that distance: to meet the dream on its own terms, as a picture. The pipeline always runs on whichever image model is currently at the frontier. I don't pin it to one vendor; I keep swapping as better models ship, because atmosphere is the thing these models keep getting better at, far more than content. A dream about a train was never really about the train. The image has to feel like the feeling. Plot illustration is beside the point.

I hope Epona becomes a quiet ritual. Something you reach for in those first soft minutes of the morning, before the day has fully claimed you. Over time, I want it to be an archive: a record of the nighttime self we so rarely get to know. There's a version of Epona I picture ten years from now, where you can look back and see what your sleeping mind was working through during the years that mattered most. That's what I'm building toward.

Your dreams are private by design. I will never sell your dream data, and I will never use your personal dream entries to train AI models. Your journal syncs via iCloud, which means it's end-to-end encrypted, and even I cannot read it. Epona earns through subscriptions. Not through your data. That's the only business model I'm willing to build on.