Hey Ya’ll! It’ been a minute since I’ve turned on the light over here, but I’m still paying the bill, lol. It’s time for me to get back to regularly posting content here. I’m mostly active on Facebook, but there’s nothing like being at “home” where I can call the shots as I see fit. So I’m flicking switch to share about my recent AI photoshoot. Yep, you read that right.
The Backstory
I think this may be my 1st post about AI content here on the blog. I’ve been creating images with AI since October ’22. I got my start on Creative Fabrica. Not long after that, I moved over to Midjourney and PlaygroundAI. Around August of 2023, I found my AI art home on Stable Diffusion (SD). That said, I still have my $60 Midjourney account so I can keep my prompts hidden, lol.
Personally, I’m fascinated with creating images in my own likeness. I tried using Picsi AI’s Insight Faceswap tool. I got ok results, but I was yearning for something more. I’m not even sure how I got turned on to SD. I think it showed up on my Youtube feed one day and I went down a rabbit hole. I ended up creating an SD 1.5 model using some of my own pics in September of 2023 and I’ve been hooked ever since.
Since then, I’ve created tons of images, trained a lora of myself in SDXL and I’ve started showing others how to fine-tune a model on their own images to create a digital twin of themselves. I’m going to write more about this in future blog posts. Let’s get into the AI photoshoot I had this weekend. I need some new headshots for some upcoming speaking gigs.
I used an SDXL model/checkpoint that I trained using Dreamlook.ai. I supplemented that with my lora file. I do that because it seems to give me the best outcomes for my likeness. I typically use the lora at a strength of 0.3 to 0.5., otherwise the images look distorted.
I got a lot of great images from the photoshoot, so I won’t have trouble with providing headshots for a while, lol. I did have to spend a little while using the inpainting feature in SD to fix the teeth so they’re presentable. SDXL still doesn’t do teeth and hands all that well on the 1st pass. I haven’t mastered inpainting hands, yet. I do pretty well with teeth.
Below are the results of my photoshoot.
The crew over on Facebook seems to like #6 for a headshot. I think the jewelry is a bit too much for professional photos. Ironically, the prompt I used was asked for a St. John Knit suit and classic jewelry pieces, lol. SD has a sense of humor. That jewelry is pretty gawdy, if you ask me. 😂
My Stable Diffusion Prompt
Prompt: photo of ukj person pretty african american female model, age 35, business headshot, wearing a black St John Knit suit, classic jewelry, gray photobackdrop, size 12, Highly detailed facial features, medium length curly wavy hair, perfect teeth, realistic lighting, subtle expressions, natural poses, authentic emotions, true to life colors, precise rendering, accurate proportions, lifelike eyes, intricate details highly detailed, lifelike, precise, accurate <lora:Meredith:0.5> .
Negative prompt: Lack of detail, pixelation, low-resolution artifacts, lack of fine textures, poor image quality, loss of detail in shadows and highlights, lack of definition in small elements, jagged edges, glasses, deformed, deformed hands, ugly, text, signature, too many fingers, floating objects, 3d, bad quality, ugly, blurry, text, watermarks, Botanical Illustration, Type Design, abstract, low detail, unrealistic, inaccurate
Style: photorealistic
Sampling Steps: 45
Sampler: Euler a
CFG scale: 7
Size: 1024×1024
Model: Meredith SDXL (my fine-tuned model from Dreamlook.ai)
My AI Photoshoot Results
This post may contain affiliate links. If you use my affiliate link, I may earn a small commission. It will not cost you any extra to use it. Thank you!
