When you create a Space, you can upload a custom image to generate a personalized avatar. The system processes your image to create an animated, lifelike digital companion - but it needs the right kind of image to work. Here is a quick guide.
What works
The key rule is: the system needs to be able to detect a face with recognizable features - eyes, nose, mouth, and lips. This applies to both photographs and illustrations. As long as the face recognition model can identify a face in your image, it should work.
- Photographs of real people - headshots, portraits, casual photos with a clear face
- Cartoons and illustrated characters - these can work, as long as the character has recognizable human-like facial features (eyes, nose, mouth, lips)
- Front-facing or slight angle - the face should be looking toward the camera, not in profile
- Good lighting - evenly lit, no heavy shadows across the face
- Sufficient resolution - high enough quality that facial features are sharp, not pixelated
- Single person - one face per image
- Clean background preferred - a simple or neutral background tends to produce better results, though not strictly required
What does not work
Will not generate an avatar:
- Faces without recognizable features - any image where the face is missing key features like eyes, lips, or a mouth. This includes robot faces without human-like features, characters with blank or featureless faces, and heavily abstract art styles
- Images with no face at all - logos, landscapes, product photos, abstract images, text-based graphics
- Faces that are partially obscured - sunglasses covering both eyes, masks covering the mouth, heavy face paint, or helmets that hide key features
- Religious or cultural face coverings (e.g. niqab, full veil): These will not generate a standard avatar because the system requires the mouth and lips to be visible for lip-sync animation. For these cases, please submit a support ticket.
Image artefacts (green screen blurring around the avatar)
Occasional green screen blurring or halo artefacts can appear around generated avatars for certain image types. If you see this, try re-uploading the image or using a photo with a plain, light-coloured background. This tends to produce cleaner avatar generation.
How will I know if my image failed?
Avatar generation is asynchronous - you will not see an error immediately when you upload. If your image fails validation, the next time you open the Space editor or test it in preview mode you will see a message prompting you to re-upload a different image or choose from the pre-generated options. This message is only visible to you as the editor; end users will never see it.
If you are unsure whether an image will work, or if you are running into issues with avatar generation, reach out to the support team and we will help troubleshoot.