How to Create Your LiveAvatar from Footage
When creating a LiveAvatar, the quality of your footage is key. It's the foundation for the entire avatar, so be sure to follow these tips closely for the best possible outcome.
We offer two options to record your footage:
Upload a Google Drive or local video file
Record with your computer's webcam
Recording Structure
Your video must be continuous with no cuts or edits, and should be a minimum of 2 minutes in total length. It must be divided into three distinct sections: Listening, Talking, and Idling.
1. Listening (15 seconds – Silent)
This opening portion is crucial for capturing natural engagement behaviors. Use facial expressions like smiling, nodding, or raising your eyebrows to show attentiveness—without speaking. This helps your avatar look actively engaged when it's not delivering lines.
2. Talking (90 seconds)
This is where you deliver your scripted message. Speak clearly, confidently, and in a natural tone. Aim for clean audio and avoid background noise. Keep your message direct and well-paced—this will directly influence the clarity and tone of your avatar’s responses.
3. Idling (15 seconds – Silent)
In this final portion, your goal is to appear present but passive. Maintain a neutral facial expression, with light occasional nods to show continued presence. Avoid dramatic expressions—this segment is used to create the avatar’s listening stance between speech responses.
Key Difference Between Listening and Idling:
During the Listening portion, you’ll show more animated engagement. In contrast, the Idling section is more static and calm, focusing on subtle presence.
Recording Environment
Background & Environment
Record against a clean, static background or a professional-grade green screen. Avoid dynamic lighting, moving people or objects, and any reflections in the background. If using a green screen, ensure it’s elevated off the floor to reduce colour spill and keep the subject at least 5 feet away to prevent shadows.
Camera Setup & Framing
Use a camera with at least 1920x1080 resolution (HD)—4K is not supported at this time for LiveAvatars due to real-time streaming latency. Frame your subject from the chest up. Avoid including hands unless absolutely necessary, and never record full-body footage (this is not supported). If hands are visible, they should remain still and relaxed at the sides or in a resting position.
Lighting Tips
Good lighting is critical for creating a clean, professional avatar. Use soft, even lighting to avoid shadows on the face and background. Two diffused lights placed on either side of the subject will provide balanced coverage. If needed, add a backlight behind and above the subject to separate them from the background, and consider warming up the light color temperature (around 4800K) to avoid pale or washed-out tones. If shadows appear beneath the hands or chin, place a soft light beneath the subject as well.
Performance Best Practices
Maintain steady eye contact with the camera throughout the entire recording.
Avoid swaying, loud breathing, or unnecessary movements.
Remove glasses, shiny clothing, or jewelry, which can interfere with background removal.
Do two takes if possible:
One with minimal movement.
One with 4–5 slow, subtle gestures that stay below the shoulders.
Talent Reminder
Before recording, coach your talent to:
Begin with a listening pose—engaged and expressive but silent.
Transition into a confident delivery for the speaking portion.
End with a still, neutral idle pose to complete the recording.
For a visual explainer on how to create the best possible LiveAvatar, please watch this short video.
How to Create Your LiveAvatar from Image
Upload a single JPEG or PNG image and get a fully interactive, real-time avatar. Head to the Avatars section in your dashboard, drop in your image, and we'll validate and process it on the spot.
Your image needs to be minimum 1080p resolution in 16:9 or 9:16 aspect ratio. The uploader validates your photo before processing—so you'll know right away if anything needs adjusting.
Image Guidelines
Recommended
Head and shoulders only, no hands—frame yourself from the chest up so the avatar has enough facial and upper body detail to animate naturally.
Face centered and fully visible—make sure nothing is cropping your face at the edges. The model needs your full face in frame to generate accurate expressions.
Looking straight at the camera—a direct gaze gives the avatar the most natural eye contact when speaking. Slight angles can cause drift in the generated output.
Mouth closed with a neutral or subtle smile—an open mouth or exaggerated expression will be baked into the avatar's resting state, which looks unnatural during playback.
One person in the photo—the model is trained to detect a single subject. Multiple people in frame will cause the upload to fail validation.
Minimum 1080p resolution—lower resolution images result in a blurry or pixelated avatar. The higher the quality of your source image, the sharper your final output.
Avoid
Cartoons, anime, or illustrations—the model is optimized for human-like features.
Hands or objects covering the face—anything obscuring your face, even partially, will degrade the quality of the avatar's expressions.
Open mouth or visible teeth—this gets locked into the avatar's neutral state and looks unnatural when it's not actively speaking.
Tilted head or side profile—the model is optimized for a straight-on angle. Significant head tilt or profile shots will produce a misaligned avatar.
Multiple people in the frame—only one face can be processed per avatar. If multiple faces are detected, the upload will be rejected.
Full-body or waist-up shots—too much distance from the camera reduces facial detail and results in a lower quality avatar.
Once you've created the best possible footage, learn how to set up your LiveAvatar in our How to Get Started with LiveAvatars article.
Note: You'll be able to train a LiveAvatar up to twice per slot a month.