Avatar v2 · guided call
Your user talks to Lily, an AI host under your brand. Nothing to film or upload. You get their avatar, their voice and what she learned about them, over the API.
Free account, no card. The call uses 30 of your 60 free tokens.

What comes back on the avatar object. The examples below are illustrative.
Cut from the best stretch of the call. 1:1 the person, ready for POST /generations.
Up to five minutes of clean speech for cloning, and clips by emotion you can pick per phrase.
Plain text, shaped by your brief. Paste it into your prompts and every script sounds like the person.
Amara makes skincare content and sells her own serum. She films UGC-style ads for her store.
POST /api/v1/capture_sessions { "context": "We generate UGC-style video ads for creators' own products.", "goal": "Who buys from them, what sells, how they talk on camera.", "language": "en", "branding": { "name": "Your brand", "assistant_name": "Lily" } } # 201 → a link for your user
One request with what your product does and what to find out. You get a link, or run the call inside your own UI with our JS SDK.
About five minutes on a laptop or phone. She interviews them by voice, follows up, and asks for a smile or a serious take if it didn't come up.
A webhook with the photo and profile seconds after the call, then the finished avatar a few minutes later. Same avatar_id as any other.
30 tokens per avatar, charged only when the call is finished. English and Russian today. Clip upload vs call
Sign up, talk to Lily for five minutes and get your own avatar. The tokens you have left make your first video.
You'll need a camera and a microphone. Laptop or phone.