Sleep onset latency of 10-20 minutes is normal, and it might be worth writing the copy around that rather than "fast". If the promise is speed, people start checking whether it is working, and that monitoring is itself arousing.
The lever with the most evidence behind it is not the bedtime ritual but a fixed wake time plus real morning light, since that sets when sleep pressure and melatonin line up. A wind-down audio can help with the cognitive arousal part; it cannot move the phase.
Two product notes: audio-only with the screen off, and a hard stop at the end. Anything still running after you doze off turns into a tracking task.
The habit most people skip, because it feels like giving up: one line telling the user that if it has not worked after ~20 minutes, get up and do something dull under low light. Stimulus control is unglamorous and it is the part of CBT-I that consistently holds up.