🔄 How the auto-failover works: The request goes to the active backend. If its ZeroGPU quota is exhausted (or it errors), this site automatically retries on the other backend — you just keep generating. Each backend has its own daily GPU quota, so you get double the allowance.
💡 Tip: keep texts short for faster generation. The first generation after a backend restart takes 2–3 minutes (model load).