An AI cup reading is only as good as the photo of your grounds. After tens of thousands of readings, the failure modes are consistent — and all of them are avoidable in thirty seconds. Here is how to take a photo the reader can actually work with.
1. Use bright, even light
Daylight from a window is ideal. Avoid strong overhead shadows and dim restaurant corners — the reader needs to see the thin trails, not just the dark clumps. If you must use artificial light, two sources from different sides beat one from above.
2. Shoot from directly above
Hold the phone parallel to the cup's mouth, not at a tea-time angle. Oblique shots stretch the rim into an ellipse and blur which zone each symbol sits in — and zone is half the meaning. Most phones show a leveling hint; use it.
3. Fill the frame with the cup
The cup's interior should fill most of the photo. Tables, hands and saucers are context the reader does not need, and a small cup in a big frame means fewer pixels on the actual symbols.
4. Show the walls, not just the bottom
The rim zone — the most time-sensitive part of the reading — lives on the interior walls. A photo that only captures the base disc loses the top of the map. If the walls are tall, a second photo from a slight angle helps; Cup Read accepts up to three views.
5. Wipe the rim, keep the grounds
Drips on the outside of the cup look like symbols. A quick wipe of the lip and outer wall keeps the reading on your actual grounds.
6. Read soon after the flip
Grounds dry out and the fine trails fade within the hour. A photo taken the next morning shows only the heavy masses — the reading loses its nuance.
7. No faces, please
Keep people, documents and anything identifying out of frame. A cup reading needs nothing but the cup — and photos are processed for the reading, not stored.
Then just ask
With a good photo, upload it and add your question — even a loose one like "what this week?" — and the reader will map what it sees, zone by zone.