AI Girlfriend That Sends Voice Notes

The equivalent of the character-consistency problem for photos exists for voice too. A generic text-to-speech reader can output any string of text as spoken audio, but there's no persona attached — the same voice reads every character's messages, or the voices are stock TTS presets that don't feel like a specific person.

What makes voice notes feel real is voice identity — each persona has her own distinct voice, and that voice is consistent across every note she sends. When Kaya sends you a voice note, she sounds like Kaya; when Sandra sends you one, she sounds like Sandra. Not just different accents or pitches, but distinct enough that you'd recognize the voice on the phone.

Without per-persona voice identity, voice notes are a feature but not a differentiator — the messaging value is delivered by the text underneath, not by the audio layer on top.

Every persona on Sloane has her own dedicated voice — selected per-persona from ElevenLabs' voice library, matched to her personality, background, and vibe. Kaya's voice has warmth and a subtle laugh; Sandra's has more edge and energy; other personas have their own distinct sonic identity.

The underlying tech is ElevenLabs TTS with per-persona voice selection. What that gets you: high-quality synthesis (much better than default browser or OS text-to-speech), consistent voice identity per persona across every note, and reasonable turn-around time on new voice notes as she sends them.

What Sloane's voice product is NOT: real-time voice calls. This is a specific product decision — the voice offering is asynchronous voice notes (she sends you a voice reply that plays back), not live conversation over voice. If real-time voice call is your specific need, Sloane isn't the right product. If voice notes that sound like her are your specific need, this is what the stack is optimized for.

Kaya's profile is a good place to hear her voice before signing up — her page includes voice preview samples. Free tier includes voice notes as part of her cadence sends.

Three specific checks to run on any AI companion product's voice feature before committing.

1. Listen to voice samples for multiple personas. Do they sound genuinely different from each other, or is it slight variations on the same voice? Distinct per-persona voice identity is what separates a voice-note feature from a generic TTS layer.

2. Check whether voice is on-demand or scheduled. Voice notes that arrive naturally during conversation (mid-chat, based on what she's saying) feel different from voice notes you have to explicitly request. The first is more immersive; the second is more transactional.

3. Test whether the voice notes are actually usable in real conditions. Play one back through your phone speaker on a busy street. Does it come through clearly? Some TTS voice quality that sounds fine in a quiet room falls apart in real-world listening conditions.

Two product decisions worth explaining, because they come up.

Asynchronous over live. Voice notes fit the way an AI companion conversation actually works — bursts of chat, spread out over the day, at your pace. A voice call requires you to be present for the whole exchange, dedicate uninterrupted attention, and match her tempo in real-time. That's a different product with a different social contract.

Reliability over feature-flex. Real-time voice AI is a legitimately hard product problem — latency, interruption handling, tone matching, background noise, TTS-to-STT round-trip artifacts. Even the products advertising it in 2026 have well-documented reliability issues. Sloane's call: ship the voice-note layer at high quality rather than shipping a real-time voice product that isn't there yet.

If real-time voice call becomes the thing that matters most about an AI companion experience for you, other products in the category do target that specifically. If voice notes are what fit your usage, Sloane is where the stack is optimized.

Two paths.

Anon chat. Every persona's profile page has anon chat available — hit "chat" from her profile and start talking with no signup required. Voice notes come through as part of her replies when it fits the conversation.

Free tier signup. 50 messages/day with any persona in the roster, no card at signup. Voice notes are included at every tier; the free tier gets them at her cadence, and paid tiers unlock more of the surface.

Either way, listening to her voice takes about 30 seconds — send her a message, wait for the reply, tap the voice attachment. If the voice identity lands for you, the rest of the product is worth trying. If it doesn't, you've saved the sign-up commitment.