Pick the right Replica (face + voice) for your AI Avatar. Stock vs your own, what you can change later, what you cannot.
Home › Help Center › AI Avatars › Choosing a Replica (the face + voice)
A Replica is the face and voice of your avatar. The same Replica can power many different personas. Picking the right Replica matters because — unlike the System Prompt — it isn't easy to change later: a "Sofia" persona feels like Sofia forever.
You have two paths:
Use a stock Replica — ready in seconds. We have a library of professionally trained faces with built-in voices. Good fit for: customer-facing avatars where the founder doesn't want to be on camera, generic explainer videos, multilingual content where finding a matching voice matters more than a specific face. Train your own Replica — record a short training video of yourself, wait around 30–60 minutes for processing, and now there's an avatar that looks and sounds like you. Good fit for: personal brands, founder-led companies, course creators, sales reps who want to scale their reach.The voice comes with the Replica. You set it when the Replica is created (for your own) or it's baked in (for stock). You can't swap the voice on a persona-by-persona basis — if a persona needs a different voice, it needs a different Replica.
If your audience speaks a non-English language and you train your own Replica speaking that language, the resulting voice will sound natural in that language. A stock English-trained Replica trying to speak Romanian will sound mechanical.Help Center · Learn · Privacy Policy · Terms of Service · Contact