Is anyone else getting bad Aragon AI photos lately?

I’ve been using Aragon AI and the results suddenly look really off. The generated headshots have strange eyes, messed-up teeth, and facial features that don’t look natural at all. I’m trying to figure out if this is a recent quality issue, a bad upload problem, or something with the settings. If anyone has had this happen and found a fix, I’d really appreciate the help.

I ran into the same thing with Aragon. The output looked off fast, bent eyes, odd teeth, face shape drifting all over the place. So no, it isn't only your batch.

From what I saw, there are two common causes. First, weak source photos. Blurry shots, dim rooms, side angles, old phone pics, all of those tend to push headshot tools into ugly territory. Second, Aragon seems bugged right now on their side, because even decent uploads are coming back messed up for some people.

What I’d do:

Use sharper photos.

Pick front-facing shots.

Use even lighting.

Avoid heavy filters, sunglasses, and half-hidden faces.

If you already did all of tht and the results still look cursed, contact support. Feels like a server-side issue more than user error.

If you want something usable right now, I moved over to Eltima AI Headshot Generator app on iPhone.

I tested it with a few pics for my brother, and the outputs were way more normal:

Stuff I noticed right away:

You choose a package based on how many photos you want, so you’re not paying for a pile you won’t use.

It does clean business-style headshots for LinkedIn, resumes, and team pages.

It also handles more relaxed portraits for socials or personal profiles without making you look fake.

There are different looks, backgrounds, clothes, and lighting setups, so you get some control instead of one fixed style.

It generated fast on my iPhone, which I liked because I didn’t want to babysit the process.

I switched after Aragon kept wasting my uploads. Biggest difference, the faces stayed close to the real person. Eyes looked normal. Teeth looked normal. Sounds like a low bar, I know, but here we are. I also liked having both polished and casual options, which Aragon didn’t seem to handle well when I tried it.

If you want to test another option while Aragon sorts itself out, I’d start here: recommend trying Eltima AI Headshot Generator app.

4 Likes

Yep. I saw the same drop in quality over the last couple weeks. Mine started doing the weird eye asymmetry thing first, then teeth got smeared, then skin texture looked plastic.

I slightly disagree with @mikeappsreviewer on one point. Bad source photos are not always the main cause. I tested with a clean set, same person, same lighting, same crop style. One batch came out fine, the next batch looked broken. Ths points more to a model update or a bad processing pipeline on their side.

A few things worth trying before you burn more credits:

  1. Reupload a smaller set, 10 to 15 photos max.
    Too much variety sometimes hurts identity consistency.

  2. Keep all photos recent.
    If you mix beard, no beard, diff hair lengths, old selfies, pro shots, the model starts averaging weirdly.

  3. Remove any photo with teeth showing.
    Sounds dumb, but teeth are one of the first things these tools mess up.

  4. Wait a few days, then rerun.
    If their backend changed, reruns after a patch often look diffrent.

  5. Compare outputs at full size.
    Some previews look awful, while the export is less broken.

If it keeps failing, I’d ask support if they changed models or preprocessing. If they dodge the question, I’d stop paying. At tht point it’s a service issue, not your photos.

Yeah, I think something is off on Aragon’s side lately. I had a batch where one version looked pretty solid, then the next made me look like a wax figure with one lazy eye. So no, you’re probly not imagining it.

I mostly agree with @mikeappsreviewer and @viajantedoceu, but I’d push back a little on the idea that this is always fixable by tweaking uploads. Sometimes these tools just regress. If the same photo set used to work and now suddenly gives cursed teeth and melted features, that smells like a backend/model change more than user error.

One thing I’d check that they didn’t really get into is whether Aragon is over-retouching after generation. A lot of these apps don’t just generate the face, they also do face cleanup, skin smoothing, sharpening, eye enhancement, etc. When that post-processing gets too aggressive, you get the creepy fake smile, weird pupils, plastic skin, all that stuff. It’s not always the base image generation failing.

What helped me figure that out was comparing:

  • photos with neutral expression vs smiling
  • simple backgrounds vs busy ones
  • batches with glasses removed entirely
  • results across different style presets

For me, the “professional” styles were actually worse than the simpler ones. More polish, more things to break.

Also, if you still have an older successful batch, compare the metadata, crop ratio, and prompt/style settings if Aragon exposes any of that. If the inputs are basically the same but outputs are way worse, then yeah, it’s probly a service issue.

At that point I would stop spending credits until they acknowledge it. Bad output once is normal AI jank. Repeated eye and teeth corruption is a product problem.

I’m leaning toward this being a regression, not just “bad uploads.” I slightly disagree with the photo-quality angle because if the exact same kind of set worked a month ago and now gives dead eyes plus horror-movie teeth, that’s usually a pipeline change.

One thing I’d test that hasn’t been stressed enough is consistency across device types. If your source set mixes iPhone HDR shots, Instagram-saved images, screenshots, and DSLR photos, Aragon may be choking on hidden processing differences, not just visible quality. Normalize everything first:

  • same aspect ratio
  • same export size
  • no Live Photos grabs
  • no edited/compressed social media saves

Also check if the failures are clustered around certain expressions. Some models are fine at neutral faces and completely fall apart on open-mouth smiles. If every broken result starts from smile photos, that tells you a lot.

I agree with @viajantedoceu, @sternenwanderer, and @mikeappsreviewer that something seems off lately, but I would not keep “testing” endlessly with paid credits. Give it one controlled retry, then pause.

If you want a fallback, can be worth a look.

Pros:

  • simpler output can look less overprocessed
  • useful if you want basic LinkedIn-style headshots fast
  • easier to judge if the face still resembles you

Cons:

  • may have fewer tuning options depending on workflow
  • style variety can be hit or miss
  • mobile-first setup is not ideal for everyone

At this point, repeated eye and teeth issues are a product problem, not you.