I’m trying to learn how to make AI photos, but I got confused after testing a few tools and the results looked nothing like what I wanted. I need help figuring out which AI photo generator to use, what prompts work best, and how to create better-looking images without wasting more time.
Making AI Photos of Yourself Without Overthinking It
I tried a few of these apps after getting tired of using the same two decent photos everywhere. The process was simpler than I expected. You upload a small set of selfies, the app builds a model from your face, then it spits out new pictures in different looks. Work profile shots, casual portraits, travel-style pics, stuff for socials, all of it.
What worked for me
The easiest one I used was the Eltima AI Headshot Generator app for iPhone.
What I did was pretty straightforward:
- I uploaded 1 to 3 clear selfies
- I used photos taken in daylight
- I picked shots from different angles
- I avoided old pics where my hair or beard looked different
After upload, the app built a personal AI model from my face. From there I picked a style pack and let it run.
The pack choices were the part I liked most. You get things like:
- LinkedIn headshots
- Corporate portraits
- Casual lifestyle photos
- Travel photos
- Social media looks
Once you pick one, it generates a batch of images. I kept the ones that looked like me and deleted the weird ones. There were a few misses, which is normal with this stuff.
Other apps I tested
If you want more options, I also looked at a couple of other apps.
GIO AI Photoshoot Generator felt more template-driven. Good if you want fast results and preset looks without messing around much.
Lensa AI is still one people bring up a lot. It leans harder into stylized portraits and editing tools, so if your goal is less ‘professional profile pic’ and more ‘cleaned-up creative portrait,’ it makes sense.
Small things that improved my results
This part mattered more than I thought.
- Use recent photos
- Don’t upload pics with sunglasses
- Skip heavy filters
- Mix in different expressions
- Use different backgrounds
- Keep the lighting clean and even
When I fed it low-quality selfies, the outputs looked off. When I used sharp photos with normal lighting, the results were way better.
My takeaway
If your selfies are clear, the whole thing is easy. Good input matters more than anything else. I had a solid run with the Eltima AI Headshot Generator app, mostly for work photos, but I used a few for personal profiles too. It saved me from setting up a full photoshoot, which was the main win for me.
You’re mixing two different things, and most beginners do.
- AI headshots of you.
- AI-generated photos from text.
@mikeappsreviewer covered the selfie-model route. I’d disagree on one part, though. You do not always need a bunch of selfies. For many tools, 10 to 20 varied photos beats 1 to 3 by a lot. More data, fewer weird face swaps.
If you want photos of yourself, pick a face-training app or service. Feed it recent images only. Same face shape, same hair, no filters, no hats. Mix close-ups and waist-up shots. Keep lighting normal. If the training photos are messy, the output gets wonky fast.
If you want scenes, products, outfits, rooms, or fake photography, use a prompt-first image model. Midjourney, Flux, Ideogram, DALL·E, and Stable Diffusion are the main lane. They differ in style:
Midjourney, strong style, pretty images, less precise.
Flux, solid realism, better prompt following.
Ideogram, good for text in images.
Stable Diffusion, most control, biggest learning curve.
DALL·E, simple to use, decent at following plain English.
Your prompt is usually the issue, not the tool. Most people write vague stuff like “make me look cool at the beach.” That gives random junk.
Write prompts like this:
subject + camera angle + lighting + setting + outfit + mood + realism cue
Example:
portrait photo of a man, chest up, soft window light, sitting in a modern cafe, navy blazer, neutral expression, shallow depth of field, 85mm lens, realistic skin texture
If you keep getting bad results, do this:
Pick one tool.
Keep the prompt format the same.
Change one variable at a time.
Save the prompt versions.
Negative prompts help too, mostly in Stable Diffusion and Flux setups:
blurry, extra fingers, plastic skin, distorted eyes, bad teeth, warped background
Best shortcut, copy a photo style you like and describe it in plain words. Lighting, lens feel, pose, background. Thats where results start looking less random.
You’re probably expecting one tool to do everything, and that’s where people get stuck. @mikeappsreviewer is talking more about face-model apps, and @espritlibre is right about prompt structure, but I’d push this a bit further: before picking a generator, decide what kind of “photo” you actually want.
Three buckets:
- You, realistically
Use a face-training/headshot app. - A made-up scene that looks photographic
Use Midjourney, Flux, DALL·E, SD, etc. - An existing photo edited
Use an editor with AI inpainting/outpainting, not a pure generator.
A lot of bad results happen because people use text-to-image when they really wanted photo editing.
Also, prompts are not magic spells. They work better when you think like a photographer, not like a poet. Instead of:
“a cool awesome instagram photo of me in Tokyo”
Try:
“realistic street portrait, standing near a ramen shop at night, wet pavement reflections, black bomber jacket, 35mm photo, direct flash, natural skin texture, candid pose”
If the face matters, use reference image + prompt when the tool supports it. That usually beats prompt-only stuff by a mile.
One thing I disagree with a little: negative prompts are useful, but beginners sometimes overstuff them and choke the image. Keep negatives short.
My practical advice:
- pick one tool for 1 week
- generate in batches
- save the exact prompts
- rate outputs by what failed: face, lighting, pose, background
- fix only one issue at a time
If hands/eyes keep breaking, zoom out less and stop asking for hyper-complex poses. AI still gets dumb in very specific ways lol.