Recording Your Own Avatar
This doc will walk you through the footage you need to record to create your custom avatar. Follow along and record just 10-15 minutes of footage so Windsor can generate your avatar.
All the footage (i.e. the model creation, consent, templates, etc.), needs to be recorded using the same microphone. The microphone cannot be a laptop mic or AirPods. A good external mic is best and they come as cheap as $50. Here ’s one we love using at $60.
To create your custom AI avatar, you will need to record about 10 to 15 minutes of audio and video footage.
First, you will need to record a short consent script on video (at the bottom of this page). This gives us permission to create your AI model.
Next, simply record yourself speaking into a camera for about 10-15 minutes.
You may record videos with your phone or your laptop. Please record at 720p or 1080p.
You may record up to 3 videos that you would like to send to recipients. Each video will be recorded as if it were to a specific individual, but in place of a name you say “Watermelon.”
e.g. (”Hey Watermelon, Thank you so much for your purchase”)
For each video recorded this way, we ask for 3 takes: Close, Medium, and Far which should be around 5, 6, and 7 feet away from the camera respectively. This will help you get the best quality.
We’ll use the footage to create a model for you and we’ll prepare the template videos you recorded using that model. Once we’re finished, we’ll let you know and you’ll be ready to go!
While we do this, familiarize yourself with the platform. Login here, create a campaign with a sample template and set up one of our supported integrations. Check out some of the other helpful docs pages on this website to help you navigate the platform.
Glad you asked. Our customers primarily use Windsor to:
These emails and videos have the highest open and watch rates. Recipients love these and typically respond to and appreciate them. It opens up new conversations and can help generate more reviews and improve brand loyalty.
This isn’t just for eCommerce brands - we even do it ourselves as a SaaS company. If you can track when a customer is on your checkout page and then doesn’t complete the transaction, a personal video that offers to help works wonders in recovering lost revenue.
If you’re collecting names and emails on your site, delight them with a personal video. Welcome them to your brand and let them know what they can expect by being a subscriber. You can also use this opportunity to have them book a call with your team - it’s the best place to do it
We typically don’t recommend using Windsor in your first email within a cold-email sequence. These emails tend to have very low engagement rates. Instead, use Windsor in follow-up emails, meeting reminders, auto-responding to lead intake forms, and so on.
If you will be sending videos via LinkedIn, they do work really well in the first post-connection message to a customer. This is because LinkedIn automatically renders a nice thumbnail when you share a link (as long as you only share the Windsor link your message).
Beyond these, it’s really up to you. Get creative with how you want to use it. If you can bring something on camera - like a sample of your (physical) product, a pet, etc. - do it 🙂
- Record all the footage necessary.
- Record the short consent script (below)
- 10 to 15 minutes of footage including your 3 template videos
- Use the same mic for all the footage
- Make sure you have separate files for 1) the consent, 2) the training reading, and 3) for each take of each template you record
I hereby verify that I, <<name>>, am currently recording this video, and I provide Windsor the permission to create an AI avatar that can replicate parts of my likeness and voice. I understand that this AI avatar can be used to edit videos recorded by me while maintaining that they will look and sound like I recorded the video myself. Since Windsor partners with Descript to generate audio that sounds like my voice, I will now provide explicit consent to Descript to use this recording. I hereby verify that I, <<name>>, am currently speaking, and that I’d like Descript to create an Overdub version of my voice. I understand that this Overdub Voice can be used to create speech that sounds like my voice.