See your dreams.
Epona turns your written dream into an image. It always runs on whichever image model is currently at the frontier, and the prompt layer on top of it is tuned for the way dreams actually speak, so the result shows the mood, not just the objects. You write what you remember. Epona gives you something to look at afterward.
It reads the dream, not the description
Most generic image models are good at the thing in the sentence and bad at the feeling behind it. Dreams are the opposite problem. The room was your grandmother's, except the ceiling was water, and the water was somehow warm. Epona's prompt layer takes that whole shape and passes forward the mood, the symbol, the setting, and the lighting as separate threads. A dream about falling through a gold-lit library does not just render a library. It renders the light, the tumble, the hush.
Always on the current frontier
I do not want to pin the product to a specific model. The state of the art in image generation moves every few months, and a commitment to a name dates badly. Epona sits on whichever image model is currently the best at the kinds of scenes dreams produce, and the prompt layer is retuned whenever we move. You will never have to read a release note to know what is under the hood. The only thing that should ever change for you is that the images get better.
Four quality tiers
Start cheap, iterate, then commit. Quick Background costs 8 credits and is enough for a first read. Standard at 15 credits and HD at 30 credits are the two you will use most. 4K at 50 credits is for the dreams you want to keep on the wall. Every tier shares the same prompt pipeline. The difference between them is resolution and fidelity, not whether the image understands your dream.
Built like cinema, not like stock
Epona was written by someone who cares about composition more than about prompt tricks. The defaults lean toward cinematic framing, soft light, and a color palette that matches the stone and gold of the app. Dreams almost always want a wider frame and a quieter color than a general model will give you, and the preset nudges the result in that direction so you do not have to micromanage it.
Regenerate until it feels right
An image is rarely correct on the first try, because no two people remember a dream the same way twice. Regenerating costs the credits of that tier, which keeps it cheap to take another pass at Quick Background and deliberate to take another pass at 4K. Most people land on the image they keep somewhere between the second and fourth try. The credits you use on drafts are the credits that let you be honest about the one you publish.
Your dream text never trains anything
The dream text you write and the images you generate from it belong to you. Nothing is added to any training set, nothing is resold, and nothing leaves the pipeline once the image is returned. The frontier model renders your dream and then forgets it. What you do with the image afterward, whether you keep it private or publish it, is up to you.
Frequently asked questions
How does Epona visualize my dreams?
- Epona runs your written dream through whichever image model is currently at the frontier, with a prompt layer tuned for the non-literal logic of dreams. The image is meant to capture the mood, not just the objects.
What are the four dream image quality tiers?
- Quick Background is 8 credits and enough for a first read. Standard at 15 and HD at 30 are the two tiers most people use. 4K at 50 is for the dreams you want to keep on the wall. All four share the same prompt pipeline.
Can I regenerate a dream image I don't like?
- Yes. Each regeneration costs the credits of that tier, so iterating on Quick Background stays cheap while 4K remains deliberate. Most people land on the image they keep between the second and fourth try.
Will my dream text be used to train AI models?
- No. Your dream text and the images generated from it stay yours. Nothing is added to any training set, nothing is resold, and nothing leaves the pipeline once the image is returned.