The [Avatars] app turns a person into a talking video. Clone yourself from a clip of you talking, or make an avatar from a photo, then give it a script or audio to speak. Your avatars are saved on this computer and show up in every project. Making an avatar is free. Videos and new looks use Selects credits.
Open the Avatars app
To open the Avatars app, click the [Apps] tab in the project panel, then click [Avatars]. The app has two tabs:
[Avatars]: your avatars, and where you make new ones
[Video]: where you make an avatar talk
Clone yourself
To clone yourself, you need a video of you talking to the camera. Selects takes your look and your voice from that one clip.
In the [Avatars] tab, click [Clone yourself].
Click [Add a clip of you], then choose [Add selected clip] (the clip selected on your timeline), [Add file], or a video from your project.
Wait while Selects checks the clip. It shows how many seconds of talking it heard.
Check the box "This is me, or I have this person's permission to clone them."
Click [Create twin].
Selects makes your twin and opens its page. Creating a twin is free.
What a clip needs:
At least 15 seconds long. Selects starts where you begin talking and uses up to one minute from there.
At least 8 seconds of talking. Your voice is cloned from it.
A video, not a photo or audio. Files can be MP4, MOV, M4V, WebM, or MKV.
Tip: Use a clip where you face the camera and talk clearly. Every video re-syncs the lips in this clip, so it's what your twin looks like on screen.
Make an avatar from a photo
To make an avatar from a photo, add a picture of the person.
In the [Avatars] tab, click the + under [Your avatars].
Choose [Add selected clip] (the frame under the playhead, or the clip's first frame), [Add current frame], [Add file] (PNG, JPG, or WebP), or a file from your project. A video from your project gives its first frame.
Selects saves the avatar and opens its page.
Under [Lock the voice], choose a [Model], then click [Choose] and pick a voice.
A photo avatar needs a voice before it can read a script. The voice models are ElevenLabs v3, MiniMax Speech 2.8 HD, and Seed Audio 1.0, and each has its own set of voices. If you change the model, pick the voice again.
You can also make a photo avatar from the [Video] tab: open the [Avatar] menu and choose [New avatar].
Manage your avatar
To manage an avatar, click it in the [Avatars] tab. Its page has:
[Name]: new avatars are called Avatar 1, Avatar 2, and so on. Type a new name and press Enter.
[Twin] or [Lock the look]: a picture from the clip a twin was made from, or a photo avatar's picture.
[Make look]: makes a new look of the same person from a description, such as "close-up in a bright office, navy blazer". Choose an image model and click [Make look]. The button shows the credit cost. New looks are kept under [Looks] for later videos.
[Lock the voice]: the model the avatar's scripts are read with. A twin can use F5-TTS (fast and low cost) or Seed Audio 1.0. Both read in your cloned voice.
[Create video]: opens the [Video] tab with this avatar.
To remove an avatar, click [More actions] (⋯) at the top of its page, choose [Remove avatar], then click [Remove]. This takes it out of your list. Projects that already used it keep its pictures, and its files stay in the avatars folder on this computer.
Make an avatar talk
To make an avatar talk, use the [Video] tab.
Click the [Video] tab, or [Create video] on an avatar's page.
Under [Avatar], choose the avatar.
Under [Look], keep the first tile (the avatar's own clip or picture), or pick a saved look. Click the + to make a new look.
Under [Speech], choose [Script] and write exactly what the avatar says, or choose [Audio] and add audio or video from your timeline, a file, or your project. With [Script], the voice model and voice sit above the script box.
Under [Video], choose a [Model].
Click [Create].
The video is made in two steps. "Reading your script" turns the script into speech, then "Making the video" makes the avatar speak it. With [Audio], Selects skips straight to the video. While it works, [Create] shows which step it's on, and stays off for that avatar until the video is done. You can leave the tab and come back. If you close the project first, the video carries on when you open the project again.
Where your avatar video lands
When the video is done, it's in your project panel under the avatar's name. If a draft was open when you clicked [Create], it's also on the timeline, at the spot where the playhead was. A script's speech is added to your project too, named after the avatar, for example "Avatar 1 voice".
Selects also adds the files the video was made from to your project: the avatar's clip or picture, and a twin's voice sample. Like the other AI apps, Selects saves the finished file to your Downloads folder, or the folder set in [Save media to].
Choose a video model
The [Video] model list depends on what the video starts from. Selects remembers the model you picked last.
Starts from | Model | Good for |
A twin's own clip | Sync Lipsync 2.0 | Lifelike lip sync on real footage |
A twin's own clip | Sync Lipsync 2.0 Pro | Finer mouth detail for close shots and beards, at about 1.7 times the cost |
A twin's own clip | LatentSync | Lowest cost, best on clear, front-facing shots |
A twin's own clip | VEED Lipsync v2 | Keeps the face and expression as they were |
A photo or a saved look | Kling AI Avatar v2 Standard | Animates the face and head of any picture, lowest cost |
A photo or a saved look | Kling AI Avatar v2 Pro | Finer faces and lip sync, at twice the cost |
A photo or a saved look | OmniHuman 1.5 | Up to 60 seconds of speech, with more body motion |
A saved look is a picture, so a twin that starts from a saved look uses the photo models.
OmniHuman 1.5 takes up to 60 seconds of speech. If your script or audio looks longer, Selects says so and [Create] stays off. A script's length is an estimate, so if the finished speech still runs over, the video stops with a message. Shorten the script and create again.
How credits work
Creating a twin or a photo avatar is free. Making a video or a new look uses Selects credits. The estimate shows just above [Create], marked "Approx.", because the length of the speech changes the cost.
If you don't have enough credits, [Create] stays off. Click [Get credits] to open your plan page.
Use an avatar in other AI apps
To use an avatar in another AI app, click the + in any box that takes a picture and choose [Avatar]. This works in [Add references] in AI image and AI video, in [Character] on the [Copy motion] tab of AI video, and in [Add references] for Seed Audio 1.0 in AI audio. The avatar goes in as its face picture.
FAQs
Can I use my cloned voice in AI audio?
Your cloned voice works inside the Avatars app only. To get a voice like yours in [AI audio], choose the Seed Audio 1.0 model in [Text to speech], then add up to 3 clips of your voice in [Add references], each up to 30 seconds. Seed Audio 1.0 takes either voice clips or one picture there, not both.
Where are my avatars saved?
Your avatars are saved on this computer, in the .selects folder inside your user folder, under avatars. Every project on this computer can use them. They aren't saved to your Selects account, so they don't show up on another computer.
Why can't I create my twin?
[Create twin] stays unavailable until the clip passes its check and the permission box is checked. The check tells you what's wrong:
"Pick a video of you talking." means the clip isn't a video.
"Needs at least 15 s" means the clip is too short.
"Needs at least 8 s of talking" means there isn't enough speech in it.
Does cloning myself cost credits?
Cloning yourself doesn't cost credits. Selects makes your twin on this computer. Credits are used only when you make a video or a new look.
What if my script is longer than my clip?
If your script is longer than your twin's clip, choose Sync Lipsync 2.0, Sync Lipsync 2.0 Pro, or LatentSync. They play the clip forward and back to fill the time.
Why can't I write a script for my photo avatar?
A photo avatar needs a voice before it can read a script. In the [Video] tab, under [Speech], choose a [Model], then pick a voice in [Choose]. You can also set it on the avatar's page under [Lock the voice]. With [Audio], no voice is needed.