Straight answers, from the people who built it.

Device requirements

• iPhone 15 Pro or newer — any iPhone with 8 GB of memory or more. The tutor is a genuinely large AI model running locally, and that memory is what makes it possible.
• ~5 GB of free storage, including the one-time model download (~4.4 GB, over Wi-Fi, with the size disclosed before it starts).
• On a device that can't run the model well, OUMA says so plainly at launch instead of degrading quietly.

Setup, in three steps

• Install OUMA (via TestFlight during the beta).
• On first run, connect to Wi-Fi for the one-time model download. The download is resumable — if it's interrupted, it picks up where it left off.
• That's it. From then on, OUMA works with no connection at all — airplane mode included.

Common issues

• “The model download won't start.” Check free storage (you need roughly 5 GB) and that you're on Wi-Fi — the first download is deliberately Wi-Fi-gated because of its size.
• “Voice input isn't responding.” Confirm OUMA has microphone and speech recognition permission in Settings → Privacy & Security. Both are processed on the device.
• “The phone feels warm during long sessions.” Running an 8-billion-parameter model is real work. OUMA monitors thermals and memory and will tell you — visibly — if it needs to ease off rather than misbehave silently.

Beta access

The private TestFlight beta enrolls a small group of families first, in waitlist order. Join from the home page — one email, no forms.

Contact

Email waitlist [at] ouma [dot] app for anything — setup trouble, feedback, beta questions, privacy questions. While we're pre-launch, replies come directly from the builder.