Hear them in their own words again
There's a difference between remembering someone and hearing them. Startover works from the actual messages they sent you — their phrasing, their punctuation, the way they joked, the way they said your name — so the voice that responds is theirs, not a generic stand-in.
Who this is for
For people whose lost loved one left a paper trail — texts, emails, voice notes, letters, social-media posts. Startover ingests that material and surfaces the voice. For people without any record, the Baimiao flow lets you describe them in 14 questions instead.
How Startover helps
Paste in the messages you have. The model learns their patterns — what they'd notice, what they'd brush off, the running jokes, the silences. Then ask what you wanted to ask. The reply will surprise you sometimes; that's what makes the rehearsal real.
When you're ready
Free during beta. Your data is yours — uploaded chat is deleted after 30 days, never used to train any model.
Common questions
Will the responses really feel like them?
With enough source material — a few hundred messages, a long email thread, a transcribed voicemail — the persona usually feels recognizable. Without source material, it will feel more like a sketch from memory.
What kinds of source material can I use?
Text messages (any chat platform), emails, letters (typed in), social-media DMs, voice messages (transcribed first), even old chat-app screenshots which are OCR-ed server-side.
Is this religious or spiritual?
Startover is a secular communication tool. It does not claim to channel a soul, contact a spirit, or substitute for any spiritual practice. It works from the words a person left behind — that's all.