Delphi Video Avatar

The primary objective of this process is to capture clear and consistent data related to your lip movements. Make sure your lips remain fully visible, well-lit, and unobstructed throughout the recording. Every aspect of the setup—from lighting to positioning—should focus on maximizing the AI’s ability to accurately track your lip movements for the best results.

This guide breaks down what to do before, right before, during, and after recording to ensure the best results.

​

Overview

• Entire process: ~15 minutes

• Follow instructions closely: even minor errors can affect avatar quality.

• Use this as a checklist for before, right before, during, and after recording.

​

1. What to Prepare Before Recording

• Background: Choose a natural setting (office, studio, or home).

  • Avoid busy or moving backgrounds (e.g., cars or people walking by).

  • Do not use green screens.

  • Ensure the background contrasts with your clothing.

• Clothing:

  • Wear clothes that leave the neck visible (no turtlenecks or high collars)

  • Avoid glasses, hats, or accessories that obscure your face.

• Equipment:

  • Camera: Use a device capable of 4K or 1080p quality (smartphone, laptop, or professional camera like SONY FX6, Red KOMODO). Do not use video call apps (e.g., Zoom).

  • Camera Stability: Keep the camera steady and at eye level; use a tripod if available.

  • Audio: Use a microphone. Avoid Bluetooth headphones like AirPods.

  • Lighting: Prepare a well-lit space. Use three-point lighting if available.

• Speech Topic: Plan for a ~3 minute impromptu speech.

  • Avoid repeating phrases or numbers

  • Practice natural pauses.

​

2. What to Check Immediately Before Recording

• Positioning:

  • Ensure the camera is stable, at eye level, and recording in landscape orientation.

  • Center your face in the frame.

  • Sit close to the camera: you should fill at least 25% of the screen.

• Lighting: Verify your face is evenly lit with no shadows.

• Audio: Check that there’s no background noise or other voices.

• Background: Confirm the background is still and distraction-free.

• Clothing: Double-check that you’re not blending into the background.

• Mindset: Relax and adopt a slightly pleasant expression for a friendly avatar.

​

3. What to Do While Recording

• This should all be filmed in one take. Do not create two separate recordings, it can cause jumps in the final avatar.

• 1. Begin with the Consent Statement: “I give Delphi consent to use this video to create my avatar.” This is mandatory, and we cannot proceed without it.

• 2. Speak for ~3 minutes.

  • This must come before the silence portion.

  • Speak clearly at a natural pace (not too fast or slow).

  • Pause for 1-2 seconds and close your mouth naturally after every couple of sentences.

  • Mistakes or mispronunciations are fine; focus on capturing lip movement.

• 3. Sit in silence for ~1 minute.

• During all parts of the recording:

  • Minimize movement as much as possible.

    • Keep your head stable: do not shake your head, nod, or otherwise move it.

    • Keep your hands out of frame: hand movement will distort your avatar.

  • Film while seated for the most natural results.

  • Never turn away from the camera.

• Expression: Maintain a slightly pleasant smile.

  • Avoid yawning, bursts of laughter, or other exaggerated emotions.

​

4. What to Review After Recording

• Video Quality:

  • Confirm the video is in 4K or 1080p quality.

  • Ensure it’s in landscape orientation and .mp4 or .webm format.

  • File size must be under 750MB.

• Consent Statement: Verify that the required consent statement is present at the beginning.

• Content: Check that:

  • Your face is centered, fully visible, and well-lit throughout.

  • Your lips are consistently visible and unobstructed.

  • There are no disruptions (e.g., background noise, interruptions, or frame misalignment).