I've been very satisfied with the safety protocols in the Adobe Express Mobile App for iPad. It's running the Adobe Firefly Image 3 Model which is trained strictly on licensed content to avoid safety issues found in open datasets.

• Minimalist UI structure
• Enterprise-level filtering
• Apple Pencil optimization It works well for toddlers since it simplifies the prompting logic. My daughter uses it on her Apple iPad Pro 11-inch M4 with no complaints tho.

I was just looking at some latency benchmarks for mobile-side inference and honestly, for a five year old, you really gotta watch those latent diffusion filters. Most apps just wrap an API, but for that iPad Pro setup, I'd probably look at Picsart SketchAI Scribble to Art iOS. It is basically built for what you're describing... just turning messy lines into actual images. It's usually around 6 bucks a week or has a monthly sub tho, so maybe check the trial before committing. One thing to be careful about with these apps:

• Make sure you use Guided Access on the iPad so she doesnt accidentally wander into any explore feeds where other people's stuff shows up.
• Stick to tools that use DALL-E 3 on the backend like Microsoft Designer Image Creator for iOS because their safety layers are notoriously strict compared to open source models.
• If she's still a bit rough with tech, a Logitech Crayon for iPad Digital Pencil is way more durable for a kid than the standard Apple Pencil. The generation time can be a bit of a buzzkill for a toddler with zero patience, so a solid 5GHz WiFi connection is actually pretty huge for keeping the feedback loop fast. If the app feels laggy, it's probably just the server-side queue being a pain... hope the birthday station turns out cool!

^ This. Also, I stumbled upon this discussion while looking for something similar for my nephew last year. Honestly, I was terrified he would accidentally generate something nightmare-inducing with my work setup. I spent hours testing stuff because I wanted that magic feel without the risk... You might want to consider Canva Magic Media for iPad. It isnt specifically a toy but their safety filters are way more aggressive than the open source stuff. I used it to let him turn his monsters into real characters. Just make sure to lock down the sharing settings first. Another one I looked at was Doodle Morph AI Art Generator. It is basically built for exactly what you described—turning scribbles into finished art. I would suggest being careful with any app that has a community tab tho. Even with filters, you never know what other users might name their stuff. Best to stick to the drawing canvas.

Yep, this is the way

Same here!

bump