Tag
← all experiments
Making a text-to-speech model decode in one pass instead of sixteen, so it runs fast on-device. With audio you can play.