A true whisper has no tone in it at all. It's breath shaped into words, with the vocal cords switched off. ASMR creators use it in short bursts. Over a twenty-minute sleep track, though, a pure whisper gets hissy and tiring, and people start straining to catch the words. For anything long I'd pick a hushed voice, one with real tone at low volume and the sense of someone close to the mic.
Closeness matters most. Record a voice near the mic and you pick up extra low-end warmth with very little room sound, so it feels like it's right beside your ear. A quiet voice recorded from across a room just sounds far away.
To hear it in a ten-second sample, put in earbuds (that's how most people listen in bed) and keep the volume low. Start with the "s" sounds. Any hiss or whistle there will repeat hundreds of times over a long track. Then listen to the gaps between phrases. In a good ASMR voice the silence feels calm, and if the gaps sound clipped or empty, the whole track will feel edited.