How to make an AI avatar influencer: one face, one voice

To make an AI avatar influencer, create one reusable digital persona and use the same face and voice in every post. On Sume, you create the avatar once, from a text prompt, a profile, or a photo, and keep its handle. Text-to-speech with the handle speaks in the avatar's voice, talking videos and lip-sync clips made with it use its face, and its public still can steer new photos as a reference image.

Sume facts come from Create new avatar, Generate avatar video, the Image API, and the Sume API reference, read on 2026-09-27; anything called current behavior is read from Sume's code. This covers making the persona, not growing an audience.

What is an AI influencer?

A virtual creator: a generated person who appears in photos, talking videos, and ads the way a human creator would. What makes it one persona rather than a string of look-alikes is consistency, the same face and the same voice every time. A reusable avatar is the starting point: you generate the identity once, then point every later request at it instead of describing the person again.

How do I create the influencer's look?

Send POST /v1/avatar-1.0/generate with an avatar_handle and one of three inputs, then poll the job; How to create a reusable AI avatar shows the request. Creation is $0.95 per avatar on API pricing, plus a 5.5% agent fee by default.

• Prompt: describe the person. In current code, Sume adds "Realistic vertical smartphone portrait, natural daylight" to your text and asks to avoid stylized illustration, so the prompt aims at a realistic portrait, not a cartoon.
• Profile: ethnicity (one of eight listed values), sex (male or female), and age from 20 to 80.
• Photo: a public HTTPS image URL.
• The handle takes 2 to 30 letters, digits, underscores, or periods, with no period or underscore first, last, or twice in a row, and the sume_ prefix is reserved for Sume's own avatars.

How do I reuse the influencer in videos and photos?

Every surface below takes the persona by reference:

How do I make photos that look like the influencer?

Read the avatar's preview_image_url, a public Sume-hosted image, and pass it as a reference image to POST /v1/images. Pick a model that takes references: the docs say ChatGPT Image 2.5 (openai/gpt-image-2.5) accepts up to 16, and a model whose catalog descriptor for input_references is {"min": 0, "max": 0} rejects them. The returned data[].url is signed, so download the image you keep. More on this in consistent character AI images.

Image 1.0 still accepts an avatar_handle directly for "face-in-still ads", but its docs mark it "Retiring soon", so new work belongs on /v1/images; see migrating from Image 1.0 and Video 1.0.

What about the influencer's voice?

The voice comes with the avatar: the API reference describes a voice preparation state on each avatar, voice.status, and text-to-speech with the handle speaks in that voice. The avatar create takes a prompt, a profile, or a photo, with no voice input.

• Talking videos speak English only in current code, where the clip prompt says "Spoken language: English only."
• For other languages, text-to-speech takes a language, and you lip-sync the avatar to that audio with VEED Fabric 1.0. Keep in mind that in current code avatar voices are cloned in English.
• Cloning or inventing a different voice happens in the Sume app, not the API: its Voices page says "Clone a voice or invent one from a prompt, then pin it in Agents chat with @."
• Current execution supports one resolved avatar per talking video, so a collab with a second persona takes separate jobs; see AI avatar conversation videos.