Import your own face: file rules and what makes a good photo

When to import instead of generate

You already have a character with a face people recognize, and you want that exact face in AvatarFactory. Importing keeps it pixel for pixel. If you only have a mood or a look in mind, generate instead: Create an avatar from scratch.

Importing your own photo is also how you clone yourself with AI, so the avatar on camera is you.

Open the import wizard

1. Open Avatars and click + New Avatar.
2. On How do you want to build it?, click Import your avatar (“Already have one? Upload its image and bring your own ElevenLabs voice”).
3. The wizard has three steps: Upload, Story, Voice.

Step 1: Upload

The heading asks Who are we importing?

1. Type the Avatar name (for example, Sarah Johnson).
2. Optional: under Link to avatar (TikTok or Instagram), paste the profile link of the account the avatar already runs (for example, instagram.com/sarah.johnson or tiktok.com/@sarah.johnson), or just its username. The field shows which platform it found; for a bare username, pick Instagram or TikTok. It lets us pull more statistics for the avatar. A link we cannot read, like a link to a single post, shows a hint and is simply not saved, so it never stops you from continuing.
3. Optional: pick a Category. Leave it as Other and we base the content on viral ideas instead of a category.
4. Under Avatar images, add the face. Drop a file on the card or click to pick one. The first image becomes the face and is used exactly as uploaded.
5. Optional: add up to five more images of the same person, by file or by pasting a link. These become extra references in the avatar’s Reference Pack. The link field only works once the face is in place.
6. Click Continue →. Name and image are required.

Under the grid you see the file name, its dimensions and its size, plus a note: we build the full reference pack from this image after you save.

File rules

• Format: PNG or JPG. WebP and iPhone HEIC photos are accepted too.
• Size: up to 15 MB per file.
• Count: six images in total, one face plus five references.

If a file is rejected, the message says why: ” is not an image we can use. Upload a PNG or JPG portrait instead.” or ” is 18.2 MB. The limit is 15 MB.” Fix the file and add it again.

What makes a good photo

• A clear, front-facing portrait, eyes open, face filling a good part of the frame.
• Even lighting, no heavy shadows across the face, no sunglasses or hands in front of the mouth.
• One person in the picture. A group shot confuses the reference pack.
• Sharp and unfiltered. Beauty filters and heavy compression carry into every scene.
• Upright. Rotate the photo before uploading.

For the extra references, add different angles and expressions of the same person. They give the scene generator a better likeness to work from.

Step 2: Story

Same as the generation wizard: a Backstory (click Ideate for me for three angles) and a Target audience (click Generate for a draft). Nothing is charged on this step. Click Continue to voice →.

Step 3: Voice

Three options in the segmented control:

• Use my ElevenLabs voice. Paste your ElevenLabs API key, then search and pick a voice from your account. Speech runs on your ElevenLabs credits.
• ✦ Design a new voice. The same three-preview audition as the generation wizard.
• Skip for now. Save without a voice and add one later from the avatar page.

Click Save avatar →. All three paths are explained in Voices.

After saving

The avatar opens on its page with your photo as the face. The Reference Pack builds in the background from that image. From here you can generate scenes, which place that exact face in new settings: Scenes and images.