Avaturn

The documentation says a developer can sign up at the developer portal and embed Avaturn free, with no limit on the number of avatars created or exported. In that route the avatar creator opens inside your app or page, the user signs in to Avaturn, and your code receives the finished GLB when the user clicks Next.

Everything that changes who owns the user flow is paid. The REST API, the web SDK for custom UI and callbacks, skipping the Avaturn sign-in page, uploading selfies directly and rendering an avatar as a profile image are listed as paid-package features. The pricing page shows one self-serve paid tier, Pro at $800 a month for up to 1,000 avatars.

The terms separate your uploaded content from the 3D model. You keep rights in photos you submit, but user 3D models are listed as Goodsize content, owned by or licensed to the company. Using an avatar requires attribution with a link to Avaturn, and commercial use is allowed only after emailing the company a link to the project.

Pro carries its own eligibility rule: it is intended for startups founded after 2021 with annual revenue below $1 million that use Avaturn in their own product. Integration for the benefit of a third party, or at $1 million revenue or more, needs an Enterprise license. The company also reserves the right to ask you to stop using an avatar at its discretion.

The privacy policy says photos and videos recorded for scanning are uploaded to Google Cloud Platform and Amazon Web Services for processing. It says they are used only to provide the scanning feature, apart from anonymized data kept for technology improvement and data annotation by service providers, where data is not linked to the account.

An avatar built from a face is personal data about that person. Scan only yourself, or someone who has agreed in writing to the scan and to the planned use, and never a photo taken from social media or a stock site. The terms forbid content containing a third party’s private information without consent and set a minimum user age of 13.

Avaturn offers two avatar types. T1 has a static face: the docs call it the most realistic and lightweight option, but face bones and blendshapes cannot animate it. T2 adds separate eyeballs and a mouth, so the face can move with ARKit blendshapes and visemes, at the cost of being a bit less recognizable.

Choose by the job. For a still render, a profile picture or a character that is only seen from a distance, T1 keeps the likeness. For lip sync, VTubing or talking characters, T2 is the only type that can move its mouth. Body presets v2023 and v2024 also differ: v2024 skeletons are almost identical across shapes, which makes shared animations easier.

Avaturn exports GLB, the binary form of glTF 2.0, which Blender and Unreal Engine can load; Unity needs the free glTFast package, and the docs advise against the FBX route there. Mixamo cannot read GLB, and the docs warn that converting GLB to FBX may still cause problems. Instead they provide a ready FBX avatar tested with Mixamo: animate that, export the animation, then apply it to your own avatar.

Downloads from the consumer hub are handled differently from app integrations. The hub sign-in page says GLB downloads are unlocked through a bot in the Avaturn Discord server, while an integrated app receives the file through the export callback. Plan for that extra step before promising a player a downloadable model.

The API reference describes four objects: a project with its own subdomain and token, anonymous users created by your backend, avatars, and sessions. Your server creates a user, opens a short-lived session, and passes the session link to the web SDK or a webview. No real user details are needed on the Avaturn side.

You must store the mapping between your account IDs and Avaturn user IDs if people should return to an existing avatar, because sessions expire after each use. Exports come either as a base64 data URL built in the browser or as a permanent HTTP link for backend use, and the choice is set in the developer portal.

1. Embed the creator on a local test page using the web quick-start code and your project subdomain.
2. Complete the avatar, click Next and save the GLB from the export callback.
3. Import the GLB into Blender and check the rig, the ARKit blendshapes and the viseme shapes.
4. Apply one Mixamo animation using the provider’s tested FBX workflow.
5. Record the steps, the file size and any sign-in or attribution prompts.

Accept a GLB that imports with a working humanoid rig, moving mouth shapes and a recognizable face. Reject a missing rig, broken textures or an export blocked behind a paid tier. No likeness or rig result is reported until the exercise is run.

For a talking 2D presenter rather than a 3D model, see HeyGen or Synthesia. To add depth motion to a flat photo instead of building a rigged character, compare Immersity AI.