The idea of virtual influencers has been around for a few years, though it’s mostly been science fiction. Over the course of 2023 and now 2024, that has changed. AI image tools like Midjourney and Leonardo have become incredibly popular (I use them every day!). But these tools struggle to generate consistent versions of people, and tools like Stable Diffusion that can generate consistent images can be difficult to use.
Enter RenderNet. This tool is clearly focused around creating consistent characters and models, so that once you’ve generated your avatar, you’ll be able to place them in multiple locations, multiple circumstances, etc.
In this article, I’ll explain what RenderNet is, and how to use it to make a virtual influencer.
Meet Ruby Reai, the Virtual Influencer
Ruby is an AI-generated avatar I created in RenderNet.ai. RenderNet is a Y-combinator-funded AI imaging startup that offers a free plan with up to 50 credits. Go ahead and sign up at Rendernet.ai (you can use the Google login for easy access).
As you can see, RenderNet really focuses on two primary features: Control Net and Face Lock. To build our virtual influencer, we’ll primarily be using Face Look, but later on, I’ll also touch on Control Net.
FaceLock for Your AI Influencer
First off, you’ll need to select your Model. For the highest quality realism, choose “Epic Realism”. This is likely the best choice for a virtual influencer. But if, for example, you were working on a comic book character that you wanted to be consistent throughout the story, you might choose another model like Night Sky or CounterfeitXL.
Your choice of model depends on your goals.
In our case, we’ll do Epic Realism. Once you’ve chosen the model, it’s time to generate your first image.
Click the face lock option, and you’ll find yourself at a screen like this one:
Consider what characteristics you’d like your AI influencer to have, and enter those details into the text box at the bottom. In my case, I started with “a 25 year old red haired woman at the beach, blurred background.”
I got a couple of images I didn’t really love, and a couple that looked good… except they were of a topless woman! Most consumer AI image generation tools don’t produce NSFW images without some wrangling, so I was suprised at this.
I liked the face that was generated, but didn’t think those images would be a good choice to use as my source image to create more. So I tried again, this time using the prompt “”a 25 year old red haired woman at a restaurant, blurred background.”
I got four images, and one of them worked very well for my AI influencer. Here’s Ruby Reai, my AI Influencer:
I figured I’d keep her at a restaurant, but have her hold a menu:
As you can see, the face lock option keeps the same face, even if the angle changes. It’s extremely high quality, and a bit easier to use than scenario.ai (which I also love).
But the hands are a bit off in many images (as you’ll see going forward).
Now it’s just time to generate more prompts!
Here are a few nighttime street shots (including one of her back, which is cool):
And here is an image from a nightclub:
Note the wonky hand in the above image. Certainly not ideal, and not what you see from Midjourney 6 and Leonardo.ai in 2024. But the consistent character trade-off is worth it!
And finally, here are a few from a photoshoot in Bali:
With RenderNet and Rubi Reai, I could pretty easily start an AI Instagram page and attempt to build a following.
And let’s face it, for some users, the surprising NSFW images would certainly be a plus in some cases as well…
What About Control Net? What Is That?
Control Net can be used to keep a subject’s position, but change the person — kinda the opposite of how Face Lock works. It allows you to use a source image of someone doing something, and apply your AI influencer to that image.
To showcase this power, I found an image on Unsplash of a woman walking through an arched hallway. Here’s the image:
I used that pose and framing as the source for my new image of Ruby Reai. I kept her Face Lock settings the same, but also added the above image as a Control Net image.
Here is the output:
Great image, just don’t look too closely at the hands. I’m sure that as RenderNet improves, the hand funkiness will go away, as it did in Midjourney.
RenderNet Credits and Payments
As I mentioned at the start, RenderNet gives you 50 free credits when you sign up. That’s good enough to get you going and messing around with the tool, but it’s not sufficient to make lots of images. Eventually you’ll need to buy credits.
The good news is that you don’t have to do yet another monthly fee like Midjourney or ChatGPT Plus! RenderNet uses a credit system, and 500 credits is only $5. Spending more than that will make the per-credit cost go down further!
RenderNet is a breakthrough technology for those looking to create consistent characters. By creatively using Face Lock and Control Net, you can create your own AI influencer and send them around the world!
Good luck, and may your AI influencer be everywhere this year!