AI talking avatars

Keep the face and voice ready for the next video

Save an avatar with photos you are allowed to use and an optional cloned voice. Its name, looks and voice stay together for your next talking video. Choose a photo and prepare speech or recorded audio, then check the available settings and credit cost. Return to the same identity when you need another script.

Available models, settings and costs are shown inside the product before you create.

Start withAn avatar photo and speech or recorded audio
Example avatar script

Welcome back to the workshop. In this lesson, I will show you how to measure the shelf, mark the holes and check that everything is level before we begin.

Talking video engineLook and audioCreate · exact credits
Workflow previewA talking video from a saved identity

How it works

Save the identity, then make it speak

  1. 01

    Create the avatar and add a look

    Open My Avatars, name the identity and select a photo from your own media. Choose a clear image where the face and mouth are easy to see. You can keep more than one look with the same avatar, which helps when a person has different outfits or presentation settings.

  2. 02

    Choose the voice or recorded audio

    Attach one of your saved cloned voices if you want to write new speech for this avatar. You can also work with recorded audio in the talking video composer. Prepare and listen to the spoken piece first so that the words, delivery and timing are ready for the video.

  3. 03

    Create a talking video

    Choose the avatar look you want, review the selected audio and check the options and credit cost shown in the workspace. Submit the video, then watch the result for facial detail and mouth movement. Return to the avatar collection when you need the same identity for another script.

Practical guidance

Build an avatar you will want to use again

A saved identity is useful when the source photos are deliberate and the voice is easy to recognize. Treat the first video as a chance to check those choices before making a longer series.

Choose the photo for speaking

A sharp, well-lit face gives the model useful detail around the mouth and eyes. Check for hands, hair or objects that cover part of the face. A photograph that looks dramatic as a portrait may be a harder starting point for a clear spoken explanation.

Organize looks by their purpose

Keep a workshop photo for lessons and a more formal look for introductions if those match the videos you make. Select the look that fits the current piece before creating. Keeping them under one avatar makes the choice easier without asking you to build a new identity for every photo.

Use a script someone could say naturally

For the workshop example, each action gets its own phrase: measure, mark, then check the level. Read the script aloud and remove phrases you would never say to a viewer. Short explanations are easier to listen to and easier to revise after hearing a speech preview.

Check the voice before the face moves

When using a cloned voice, listen for names, emphasis and pauses before creating the talking video. If you use an existing recording, trim unwanted silence first. The chosen audio drives the spoken performance, so fixing it early gives you a better basis for judging the finished clip.

Review each new video on its own

Using the same saved identity gives you the same source choices, but generated results can still vary. Watch for changes in facial features, mouth shapes and the edges of the photo. A successful first video is a useful reference for your review, rather than a guarantee about the next one.

Questions

Frequently asked questions

What is saved inside an avatar?

An avatar keeps a name, one or more photo looks and an optional cloned voice from your account. You choose the look when making a talking video. This lets you reuse the identity without selecting an unrelated face and voice from scratch each time.

Can one avatar have different photos?

Yes. You can save several looks with an avatar and choose the photo for each video. Use images you are allowed to use and check that the face is clear in each one. Different lighting, framing and head angles can affect how a generated talking video looks.

Does Luma invent a new face when I create an avatar?

The avatar collection starts with photos you select. Its job is to keep those looks and an optional voice together for talking videos. If you want a different source image, create or prepare that image first, then add it as an avatar look.

How does this relate to Lip Sync?

Both workflows create a talking video from a still photo and audio. My Avatars adds a saved identity with reusable looks and a voice choice, which is helpful for recurring videos. Lip Sync is a direct starting point when you already have a photo and a recording for one piece.

Start with the idea you have now

Add a photo to your avatar collection and choose the voice or recording for your first video. Review the speech and video before using the finished piece.

Create your avatar