This is really cool. Probably you have a beefy gpu. If you chose the schedulers wisely you can get high speed results with less inference steps.