How to Record a Custom Video for the AI Autopilot Inbound Agent's Video Avatar

Last updated: June 25, 2026

Warmly's AI Autopilot Inbound Agent can show a lifelike video avatar of you or your team on your website — powered by Tavus. To create your avatar, you'll record a short 1-minute training video and send it to your Warmly CSM. We handle the upload for you.


Before You Start

Make sure you have:

  • A laptop or desktop with a built-in camera (or external webcam)

  • A quiet, well-lit room

  • A plain background (wall, solid backdrop)

  • About 10 minutes


Setup Checklist

Appearance

  • No glasses, hats, or jewelry — accessories interfere with avatar generation

  • Wear clothing that contrasts with your background (avoid a white shirt against a white wall)

  • Keep hair behind your shoulders, off your face — no bangs or loose strands covering your face

  • Avoid high collars or turtlenecks — your neck should be clearly visible

Camera & Framing

image.png
  • Position your camera at eye level (prop your laptop on books if needed — don't look down at it)

  • Sit waist-up in frame: head, shoulders, and upper chest visible

  • Your face should fill at least 25% of the frame — not too far back

  • Sit at least 3 feet from the camera, as if you're on a normal Zoom call

  • Do not use a browser-based recorder — use QuickTime (Mac), the Camera app (Windows), or Loom

Lighting

  • Face a window or soft lamp — light should hit your face from the front, not from behind

  • No harsh shadows across your face

  • Consistent lighting throughout — avoid recording near a window where cloud cover will change the light

Audio

  • Use your device's built-in microphone — do not use AirPods, wireless earbuds, or a standalone high-end mic

  • Turn off any audio effects: noise suppression, EQ, spatial audio

  • Record in a quiet room — no fans, AC noise, background music, or echo

Background

  • Simple, still background — a plain wall or office backdrop works great

  • No other people or movement visible in the background


What to Record

Your video is exactly 1 minute long, recorded in one continuous take (no cuts or edits). It has two parts:

Part 1 — First 30 Seconds: Speak Freely

Talk naturally, as if you're greeting a prospect who just landed on your website. Some ideas:

  • Introduce yourself and what your company does

  • Explain how your product is different from competitors

  • Invite them to ask you anything

Tips for this segment:

  • Speak clearly with natural enunciation — you want your lips and teeth to be visible

  • Keep your head and body mostly still — minimal head turns, no hand gestures

  • Don't block your face or mouth with your hands

  • Look into the camera lens, not at your screen

Part 2 — Last 30 Seconds: Silent & Still

Stay in the exact same position with the same expression. Just blink and breathe normally. This is the "listening" state — the avatar uses this footage when the prospect is talking and the avatar is on screen but not speaking.

Tips for this segment:

  • Keep a neutral, relaxed expression — closed lips, no smiling or frowning

  • Stay as still as possible

  • No fidgeting, no looking away, no movement


Technical Requirements

Requirement

Spec

Resolution

1080p minimum (4K preferred)

Frame rate

25 FPS or higher

File format

.mp4 (H.264 + AAC) or .webm

Max file size

750 MB

Length

Exactly 1 minute

Takes

One continuous shot — no cuts


How to Send Us Your Video

Once you've recorded, share the file with your Warmly CSM — either by:

  • Uploading to Google Drive and sending the link

  • Sending the file directly via email or Slack

Your CSM will handle uploading it to the platform and will let you know when your avatar is live (typically within a few days of submission).


Common Mistakes to Avoid

Mistake

Why It Matters

Wearing glasses

Reflections and lens distortion degrade the avatar

Camera angle too low (looking down)

Unnatural angle makes the avatar look odd on screen

Moving hands or head frequently

Creates visual artifacts in avatar output

Using AirPods or noise-suppression mics

Alters audio in ways that hurt lip-sync accuracy

Recording in a browser tab

Browser recorders compress quality; use a native app

Background movement or noise

Reduces avatar fidelity

Stopping mid-take or editing the video

Must be one continuous unedited shot


Questions?

Reach out to your Warmly CSM or email [email protected].