I’ve been using this tool at Tier 3 and have generated around 60 images so far. Some of these came from the onboarding process after training my model, while the rest I created myself. Overall, I really like the product and think it has a lot of potential, but there are a few things worth knowing before you dive in.

Credit System: Too Tight for Comfort
The biggest drawback is the credit system. Training a model consumes a large number of credits upfront. If I had only purchased Tier 1, I’d have nearly exhausted my credits just from training and generating those 60 images. Even Tier 3, which claims to support up to 8 models, feels tight. Training one model and generating a decent number of images used almost 2,000 credits. Technically, I could train more models, but that would leave fewer credits for actual image generation. It’s not a dealbreaker, but definitely something users should be aware of.

Image Quality: Impressive but Not Perfect
The output quality is generally strong. Many of the images I’ve generated are already in use across various projects. The model captures facial features fairly well, though it occasionally misses the mark. Sometimes the face is slightly off, or the body type doesn’t align with reality. It also tends to default to idealized or very fit versions of people, which can’t be corrected through training alone. You can nudge the result with prompt tweaks, but the training doesn't fully account for body shape.

One big flaw is the handling of text. Any generated writing or signage in images tends to be unreadable AI gibberish. Also, like with many AI tools, weird details, such as extra fingers or awkward hand poses, can slip through. That said, most of the time, the results are on point.

Features and Flexibility: Some Nice Touches
One feature I appreciate is the base prompt enhancer. It helps refine your input to get better results without needing to master prompt engineering. The built-in expert curated presets are hit or miss, though. While some are inspiring, many of the Tier 3 presets feel unnecessary and aren’t something I would use.

Aspect ratio flexibility is locked behind Tier 2 and Tier 3, which makes Tier 1 feel even more limiting. On the plus side, once your model is trained, image generation is fast. Enabling face enhancement does add a bit of processing time, but it's still reasonable.

Final Thoughts
The longest wait comes during the model training process. But honestly, it was nice to walk away and get notified by email when it was done. After that, everything runs pretty smoothly.

I give it 4 tacos out of 5. With more generous credit allocations or rollover for unused credits and body training, it would easily be a 5. It’s a promising tool, and I hope they keep improving the pricing model to match its potential.