Get Accurate AI Headshots: How to Create Your Flux LoRA

Tired of using outdated or irrelevant headshots? Wondering how to create a mix of professional-quality headshots without hiring a photographer?

In this article, you'll discover how to generate realistic AI headshots that look like you for use in marketing, social media, and business applications.

Why Marketers and Entrepreneurs Should Care About AI Headshots

AI-generated headshots offer remarkable flexibility for anyone who needs custom images for social media, websites, and other marketing channels. 

The cost is minimal—typically a few dollars for a set of images, compared to hundreds or thousands for a professional photo shoot. And while early AI-generated images often looked artificial or cartoon-like, today's technology has dramatically improved. Though not every generated image will be perfect, you can usually get several usable, realistic images in just a few minutes.

For example, Jonathan Mast shared that he'll be speaking at a college campus but didn't have any pictures of himself at a campus he could use to promote the event. Within about three minutes, he used an AI headshot generator to create a dozen images of himself at a college campus dressed in a navy blazer, a blue oxford, and a baseball cap.

This quick turnaround allows you to create context-specific images that match your marketing message more effectively than generic headshots.

Other practical applications include:

• Creating relevant images for social media posts
• Generating specific poses or expressions for marketing materials
• Placing yourself in different locations for event promotions
• Wearing different outfits without needing multiple photo sessions
• Producing professional headshots without hiring a photographer
• Updating your LinkedIn profile with a fresh, professional image
• Creating consistent “About the Author” photos for guest posts
• Developing varied imagery for course materials
• Generating custom thumbnails for video content
• Creating promotional materials for speaking engagements

How to Create Professional AI Headshots With Flux LoRA: A Step-by-Step Guide

#1: Gather Your Training Images

To start, you'll need to gather at least a dozen photos of yourself to train the AI, but two dozen is better. These can be current headshots, candid photos, or selfies and should include a variety of photos with the following:

• Head and shoulder shots
• Torso-up images
• Some full-body shots
• Different angles and expressions
• Various outfits

If you wear glasses or have distinctive features like Jonathan's white beard, include images that clearly show these characteristics. However, if you often wear a type of accessory that changes, such as a baseball cap, gather pictures without this accessory shown because it's easier for AI to add an accessory than to remove it.

Lighting Considerations

For consistent results, make sure your face is clearly visible in all images and try to include shots with similar lighting conditions.

Bright sunlight might cause squinting or overexposure, and hard shadows can confuse the AI about your facial features, so don't stand with bright light sources behind you—photos taken near windows with natural daylight work best.

Image Quality Requirements

• Resolution: Use the highest resolution available (at least 1000px on the longest side)
• Size restrictions: While there's no strict file size limit, each image should ideally be under 10MB
• Format: JPG, PNG, and HEIC (iPhone) formats all work well
• Cropping: Leave some space around your face; don't crop too tightly

#2: Access Flux

To help you access Flux through the developer (Black Forest Labs), Johnathan recommends going through the website fal.ai. We created a shortcut redirect to get you there: socialmediaexaminer.com/lora.

When the site opens, click on the Sign Up button, then click on Continue with GitHub; if you don't have a GitHub account, you can easily create one for free as part of this workflow.

After creating your account, you'll need to add funds to pay for model training and image generation on Flux:

1. Navigate to Account (found in the top right menu), then Billing and Add Credits.
2. Add a payment method and fund your account with $10-20 to fund your image generation (each image costs about 3-5 cents, and training the model costs about $2).
3. Confirm your payment and wait for the funds to appear in your account; this usually happens immediately.

Once your account is funded, click on the Explore tab. In the search box, type “Flux-LORA-fast-training” and click on the model.

#3: Train Your Custom Flux LoRA AI Model to Generate AI Headshots

Once in the Flux-LORA training model, look for the Input box to upload your collection of training images, either one at a time or as a zip file.

Next, choose a trigger word–a unique term that you'll use in prompts to tell the AI to use your likeness during new headshot generation. This should be a unique term not often used in your prompts. Jonathan uses his name, “JonathanMast.”

Finally, click on Start to begin the training process, which takes about 10-15 minutes and costs about $2

The system will create a LoRA model (Low Resonance Adaptation) of your face. When the training is complete, the model will appear in the training history box on the right side of the screen.

#4: Use Your Flux LoRA AI Model

After your model is trained, you can start creating images.

Develop Your Image Prompt

Jonathan recommends using a custom GPT called Felix Image Prompter (available at socialmediaexaminer.com/felix) to help generate effective prompts for Flux.

Simply tell the custom GPT what type of image you want to create, and it will give you five image prompt iterations from which to choose. Be specific about these elements to tell the custom GPT what you want:

• Facial expression (smiling, serious, thoughtful, etc.)
• Clothing (be specific about colors and styles)
• Setting or background (describe it in detail)
• Lighting conditions (soft, natural, dramatic, studio, etc.)
• Camera angle or composition (close-up, profile, etc.)

You'll also want to specify any distinctive features you want to appear in the image, such as “with glasses.” Even if you're wearing glasses in all your training images, if you don't remind the AI that you want it to include glasses, about 80% of your images will generate with glasses, and about 20% won't.

Next, choose a prompt and include your trigger word where it makes logical sense, typically replacing references to a person. For example, instead of “a man standing on a college campus in Arkansas,” you would write “JonathanMast standing on a college campus in Arkansas.”

Here's an example of a detailed prompt that would likely produce good results:

JonathanMast with a long white beard and glasses, smiling confidently, wearing a navy blazer, standing in front of a red brick university building, soft natural lighting.

Generate New Headshots Wtih Flux LoRA

Return to the Flux model, click Run Inference, and enter your prompt. Then, choose the desired aspect ratio. 

Next, consider the Guidance Scale. Ranging from 0 to 35, this scale controls how much creative freedom the AI has when it generates an image. Set the scale too low, and the images will look less like you. Too high, and your images might look less realistic. For example, your skin may look plasticky. Jonathan recommends starting with a setting between 10 and 15 and experimenting to find the right balance between creativity and accuracy.

Then, choose the number of images you want to generate. Remember, each image costs about 3-5 cents. 

Finally, choose the output format.

The generation process takes about 30-60 seconds. Don't worry if some images don't turn out perfectly – Jonathan says he typically discards about half of what he generates.

Be sure to download and save your favorite images immediately after generating them. While they remain available in the Requests tab history, it's a best practice to save your favorites right away to avoid any potential issues with the platform's storage limitations.

Troubleshoot Low-Quality Results 

If you're not getting good results, try:

• Being more specific in your prompts about what you want
• Uploading more varied images of yourself
• Checking that your training photos have good lighting and clear face visibility
• Checking the placement of the trigger word in your prompt
• Adjusting the guidance scale to find the optimal setting

Other Notes From This Episode

• Connect with Michael Stelzner @Stelzner on Facebook and @Mike_Stelzner on X.
• Watch this interview and other exclusive content from Social Media Examiner on YouTube.

Where to subscribe: Apple Podcasts | Spotify | YouTube Music | YouTube | Amazon Music | RSS

✋🏽 If you enjoyed this episode of the AI Explored podcast, please head over to Apple Podcasts, leave a rating, write a review, and subscribe.

---

Stay Up-to-Date: Get New Marketing Articles Delivered to You!

Don't miss out on upcoming social media marketing insights and strategies! Sign up to receive notifications when we publish new articles on Social Media Examiner. Our expertly crafted content will help you stay ahead of the curve and drive results for your business. Click the link below to sign up now and receive our annual report!