Step three on Gymini is the look, and it finally looks like the thing I picked. A grey clay figure behind a countdown you can read with the phone on the floor. One green on a dark room, and nothing else is lit. Dark only, by the way. There is no light theme. The About screen says it in one line: Gymini is dark, it's a gym.
All 173 figures are rendered now, one per movement, for about three dollars. I nearly spent seventy. I'd quoted the render on the wrong model, and the thing that saved me was a script I'd written a month ago for something else, which batches the same job at half price on a cheaper model. Twelve as a pilot to check the clay held, then the rest in one go.
What broke: Pause froze the whole app for thirty seconds, and every test was green. Asking the speech engine "are you still talking" right after telling it to stop blocks the main thread until the engine feels like answering, voice guidance on or off. One line fixed it. Measured, that's 30.1 seconds down to 26 microseconds. The tests never saw it because they use a fake voice. The real one is the one that bites.
Next is the platform: reminders, widgets, and bringing your history over from the old app without touching it.
Posted on X the same day. Replies there, or email votiquellc@gmail.com.
