Be in two places at once
Build a character with your face and your voice, then let her talk to your audience one to one — while you are asleep, filming, or simply off the clock.
One character, every format you already publish in
A character that is recognisably yours
Generate a face from a prompt or a reference, then lock it. Every later image and every session frame keeps the same person rather than drifting into a stranger.
ComfyUI · appearance lockYour voice, from one clip
Clone from a single reference sample. No studio time, no scripted read, no dataset to assemble.
GPT-SoVITSOne-to-one conversations at any hour
Your character can hold a live video conversation with a fan while you sleep — hearing them, answering in your voice, with her face moving as she speaks.
Live sessionsStills for promotion
Generate images of the same character for social, thumbnails and previews, without another shoot.
Image generationClips without filming
Drive a lip-synced clip from audio, or generate short video from a prompt or a still.
MuseTalk · WAN 2.2Privacy, if you want it
Work through a generated face and a synthetic voice and keep your own out of it entirely. The character is the performer.
Fully synthetic option
An afternoon, not a production
Build the character
Generate a face or upload a reference, set the personality, and lock the appearance so it stays consistent.
Give her your voice
Upload one clean clip. It is transcribed for you and registered as a voice bound to that character.
Open a session
Start a live conversation yourself to see how she holds up, and adjust the personality until she sounds like you.
Put her in front of people
Use the dashboard directly, or embed sessions into your own site with a session token so your audience never leaves it.
She takes the night shift
What fans pay for is attention, and you only have so many hours of it. She has the rest — the same face, the same voice, one to one with whoever turns up.
- Embedded on your site. Sessions publish into a WebRTC room your page joins with a short-lived token — the audience, and the relationship, stay yours.
- No per-message meter. Self-hosted models mean a long conversation is a hardware question, not a per-token bill that grows with affection.
- She never breaks character. Uncensored weights — there is no refusal layer to shatter the moment mid-conversation.
- Always the same her. Appearance lock and a cloned voice: session forty looks and sounds like session one.
You’re up late again. Come here — tell me what kept you.
What happens to your face and voice
Your reference clip and your character stay in the deployment you are using. They are not pooled into a training set, not shared between accounts, and not sent to any third-party AI provider — every model runs in a container rather than behind someone else’s API.
You can delete a character or a voice, and the records go with it. Deleting your whole account goes further: your reference clips, portraits and generated media are unlinked from disk as well as removed from the database, and the operation tells you what it removed rather than just claiming success.
Clone your own voice, or one you hold documented permission to use. Cloning someone else’s without their agreement is grounds for termination under the acceptable use policy.
Before you upload anything of yourself
Do I have to use my own face?
No. You can generate a fully synthetic character and keep your own identity out of the product entirely — the character is the performer. Using your own face and voice is the option, not the requirement.
How much audio does the voice clone need?
About ten seconds of clean reference. The engine is zero-shot, so there is no training run and no dataset — the clip is transcribed for you and registered as a voice bound to your character.
Do you take a cut of what fans pay me?
No. Pricing is a subscription plan, not a revenue share — there is no mechanism in the platform that meters or takes a percentage of what you charge. Fans pay you, on your platform; you pay for the plan.
Can fans talk to her on my own site?
Yes. Sessions publish into a WebRTC room. Your server creates the session and hands the browser a short-lived token, so she appears on your page and your audience never leaves it.
Who owns my character and my voice?
They stay in the deployment you are using — not pooled into a training set, not shared between accounts, not sent to a third-party AI provider. Deleting a character, a voice or your whole account removes the underlying files as well as the records.
Can I clone a voice that is not mine?
Only with documented, revocable permission from the person it belongs to. Cloning anyone else’s voice without that is grounds for termination under the acceptable use policy.