Feedback on new AI tool for dealer photography?

[LucasMachado]

This is just a GPT wrapper.

 

[jasonlbaptiste]

This is just a GPT wrapper.

Not at all. We have our own proprietary algorithms that work in conjunction with other models. We are not even using ChatGPT or OpenAI for this.

 

[ollieheart]

I’ve been in this space a while and honestly this looks solid from a speed and ops point of view. Anything that helps dealers get cars online faster is always a win. This is something similar I’ve come across recently with Spyne… same kind of setup, upload your photos and clean them up quick. Works pretty well when your inventory shots are inconsistent or you’re trying to scale. Only thing… I’ve seen overly cleaned-up photos backfire a bit. When everything looks too perfect, people start questioning it. Shoppers usually trust what feels real. If it’s just enhancing lighting or fixing backgrounds, that’s great. But if it starts hiding imperfections, that’s where it can create problems later. Overall feels like a solid tool for speeding things up and fixing rough photos, just not a full replacement for real photography yet. Curious how it handles stuff like damage, reflections, or tricky lighting.

 

[carsalesphoto]

What I keep coming back to in these tools is whether the car still looks like the car that will show up on the lot.

Buyers notice when panels, reflections, or damage get “cleaned up” by the model. If a scratch or dent disappears in the listing shot, that’s not merchandising anymore — that’s a trust problem.

The workflows that feel safest to me are the ones that treat background and lighting as the edit surface, leave the vehicle geometry alone, and make you compare the finished frame to the original before it goes live. Convenience is nice; being able to reject a bad render matters more when you’re putting your name on the listing.

Curious how others here are QA’ing AI outputs before they hit the site — side-by-side review, manager sign-off, or something else?