Talking avatars and saved characters in Claude and ChatGPT

Yes. Connect Makemancer and Claude, ChatGPT or Codex can make a photo talk with HeyGen Avatar 4: give a face and either a script, read in a stock voice, or a finished recording, and the lips follow the speech. Save the face and voice on a Character and reuse them in every clip. A 10-second clip from a recording costs $2.

Bring
A portrait or a character idea, and a script
Get
A talking video, and a Character to reuse
From
80 credits · $0.80

Made with it

Our own results, first takes. Each caption says what it took.

The sheet does the remembering

Wren Adair is a made-up creator for a made-up sunscreen, Dayglaze SPF 50. Her reference sheet (front, side and two expressions) is saved on her Character, and every new image of her is made from it, so the face and outfit hold from one frame to the next.

Reference sheet for Wren Adair, a fictional creator: full-body front and side views plus two expressions
InReference sheet for Wren Adair, a fictional creator: full-body front and side views plus two expressionsNano Banana 2 · 2K · 4:3 · 12 credits
Wren Adair creator portrait, made from her saved Character
ResultWren Adair creator portrait, made from her saved CharacterNano Banana 2 · 2K · 4:5 · 12 credits
Wren Adair holding a Dayglaze SPF 50 tube on a sunny balcony, a campaign still from her saved Character
ResultWren Adair holding a Dayglaze SPF 50 tube on a sunny balcony, a campaign still from her saved CharacterNano Banana 2 · 2K · 4:5 · 12 credits

Her face and her voice, talking

Her voice was designed from a short text description and saved on the same Character. A vanity frame made from her sheet and a script read in her voice went to HeyGen Avatar 4 in its expressive style, which made the 12-second clip with the lips following the recording.

Wren Adair at a bathroom vanity with a Dayglaze tube, framed for a talking-head video
InWren Adair at a bathroom vanity with a Dayglaze tube, framed for a talking-head videoNano Banana 2 · 2K · 16:9 · 12 credits
InWren Adair's designed voice reads her talking-head scriptHi, I'm Wren. My whole morning routine? Cleanser, moisturizer, and one pump of Dayglaze. No white cast, no sticky shine. Sunscreen you'll actually wear.Fish Audio Drama 3 · 2 credits
ResultWren Adair talks to camera in her own designed voice, HeyGen Avatar 4HeyGen Avatar 4 · 720p · 16:9 · 242 credits

A presenter from one photo and a typed script

No recording at all: a fictional presenter made as one image, then two short scripts typed into the request and read in a HeyGen stock voice. The first uses the default stable style, the second the expressive one; compare how much they move.

A fictional presenter at his desk, the avatar image
InA fictional presenter at his desk, the avatar imageNano Banana 2 · 16:9 · 12 credits
ResultThe presenter speaks, HeyGen Avatar 4HeyGen Avatar 4 · 720p · 100 credits
ResultThe presenter signs off, HeyGen Avatar 4 expressiveHeyGen Avatar 4 · 720p · 80 credits

How to do it

  1. Connect Makemancer

    Add Makemancer to Claude, ChatGPT, Claude Code or Codex. The steps for each client are on the connection page.

  2. Make the character, or bring your photo

    Describe a fictional host and approve a character sheet, or upload clear, front-facing photos of yourself. Save the result as a Character, and add a voice: a stock one, one designed from a description, or a clone of your own.

    Create a fictional presenter for my cooking channel: a cheerful woman in her thirties with a striped apron, in a bright kitchen. Show me a character sheet first, then save her as a Character called "Chef Ada" and design a warm, upbeat voice for her.
  3. Give it something to say

    Write the script, choose the framing and the style (stable or expressive), and ask for a short test before a long take. Check the likeness and the lips, then make the rest.

    Make Chef Ada say this to camera, 16:9, expressive: "Today: the only pancake recipe you will ever need." Read it in her saved voice and animate her with HeyGen Avatar 4. Tell me the cost first.

Which model to use

ModelBest forPriced forCredits
HeyGen Avatar 4video/heygen-avatar-4A face photo and a script or recording to a talking clip, 720p or 1080p, stable or expressive.Opens after your first credit purchaseper input120$1.20
H3 Max Lip Syncvideo/minimax-h3-max-lip-syncOne still portrait animated to a recording from 5 seconds to 15 minutes, 480p to 2K.8 s audio input80$0.80

Build the character first

A talking clip is only as steady as the face you start from. Approve a character sheet before any video: front and side views, a couple of expressions, and the outfit you want to keep. Saved on a Character, the sheet travels with every request, so a portrait, an ad still and a talking clip a week later all show the same person. Review the likeness in each new result anyway; small drifts are cheaper to catch on an image than on a video. The sheet and each new frame cost $0.12 an image with Nano Banana 2 at 1K, a designed voice costs $0.02 for three options to choose from, and the Character's voice reads 1,000 bytes of text (about 1,000 characters of English) for $0.10.

Script or recording

With a script, HeyGen Avatar 4 reads it in one of its stock voices, which is the quickest way to a test. With a recording, the clip follows your exact words, pace and pauses, and the voice can be anything: a voiceover, a voice designed for the Character, or your own cloned voice. For a still portrait and a long recording, H3 Max Lip Sync animates up to 15 minutes in one go.

Faces you may use

Use fictional characters, yourself, or people who have agreed to it. Do not make a real person say something they did not say. For an existing video rather than a photo, lip sync changes what a speaker says and character swap changes who is on screen.

Questions

How do I make a talking avatar?

Connect Makemancer, give the assistant a face (a photo, or a character it designs for you) and a script or a recording. It quotes the cost and returns a talking video with a permanent link. Save the face and voice as a Character to make the next one by name.

Can I use my own photo?

Yes. A clear, front-facing photo with even light works best, and a few photos make a better reference sheet. Use photos of yourself or of people who agreed to it, never someone else without permission.

Can the avatar speak in my own voice?

Yes. Clone your voice from a short recording of yourself, save it on the Character, and every script is read in it before the video is made.

How do I keep the same character across videos?

Save it as a Character. The face sheet and the voice go along with every request that names the Character, so each portrait, ad still and talking clip starts from the same reference instead of a fresh guess.

What does a minute cost?

HeyGen Avatar 4 is charged per second of speech: a 10-second clip from a recording costs $2, so a minute costs six times that. HeyGen Avatar 4 opens after your first credit purchase; H3 Max Lip Sync is open from the start.