Type the sound you want to drift off to (tropical birds, waves on a shore) and Somnus would build it on the device, for as long as you needed. No stock loops, no streaming. Just the sound in your head, on tap.
That was the idea. On a phone it fell flat. Generating the audio took an age, and what came back sounded less like the ocean and more like a dying fan.
What killed it
On-device audio generation isn't ready. Too slow to be usable, and the output was rubbish no matter how I prompted or buffered it. Text and vision on a phone are getting good. Sound, for now, isn't, so I put it to sleep.