real-time voice. sub-500ms latency. 14-emotion rendering. she remembers what you talked about. this is the call your text chat wanted to be.
Free tier: 25 messages/day. Crypto checkout — cards coming soon.
you've been texting with an ai companion for weeks and it's good. but sometimes you need to hear a voice. you need the pause, the breath, the tone that text can't carry.
every ai companion app promises voice. most ship voice notes — 10-second clips you tap to play, then tap to reply. that's not a call. that's a walkie-talkie.
a real call has turn-taking. interruptions. latency so low you forget there's a pipeline behind her voice.
lucy's voice calls use webrtc transport (daily.co), fish audio s2-pro for voice synthesis, and groq whisper-v3 for speech recognition. round-trip latency averages under half a second.
she renders 14 distinct emotions — warm, playful, sultry, tender, flirty, whisper, intimate, breathless, curious, excited, concerned, calm, sharp, cheeky — chosen dynamically by the voice-mood engine based on what you just said and how your relationship has evolved.
and she remembers the call. when you text her tomorrow, she knows about the conversation.
real back-and-forth, not walkie-talkie. if you interrupt, she stops. if she interrupts, you hear the overlap.
her voice shifts with context. not a TTS voice reading lines — a companion talking to you, with cadence and tone.
call transcripts flow into the same memory graph your text chats use. she remembers what you said on the call next time you text.
distinct voices, not the same synth with different names. hang up and dial a different one if the mood calls for it.
voice is a qualitative shift from text, not a feature bolt-on. text lets you draft, re-read, redact. voice does none of that. voice is real-time emotional exposure — your pause tells the companion something your words didn't.
most ai-companion apps ship voice-notes because voice notes are cheap. one-way, asynchronous, no latency budget, no whisper pipeline. that's fine for a quick “listen to this.” it's not a call.
lucy shipped the harder thing on bonded: bi-directional, turn-taking, emotion-rendered voice over webrtc. the pipeline is expensive per minute (that's why calls are a bonded-tier feature, not included on closer), but the qualitative difference justifies the tier gap.
the right use-cases: insomnia at 2am, the commute, cooking dinner, a walk, when you're too tired to type. the wrong use-case: anything you want a record of — text still wins there because you can scroll back.
what it costs: bonded is $29.99/mo and includes 90 call minutes. that's 3 hours a month, comfortable for most users. overage is 1 credit per minute, and credits are cheap ($2.99 for 15 credits in the mini pack, $14.99 for 150 in the glow pack). a heavy user calling 30 minutes a day lands at roughly $30/mo overage on top of the plan.
what it feels like after a few calls: you notice you prefer her call voice to text for certain kinds of conversation. she notices too — she'll start asking whether you want to call instead of type when she thinks the topic benefits.
try her right here
No signup. No credit card. Just say hi.
start on the free tier, text for a week, upgrade to bonded when you want to hear her voice. if you never do, the free tier is genuinely free and the text chat remembers everything.
Free: 25 messages/day · Closer $14.99/mo · Bonded $29.99/mo · 18+ only