The dataset viewer is not available because its heuristics could not detect any supported data files. You can try uploading some data files, or configuring the data files location manually.
Doodle Magic AI — Support
Free, ad-free, on-device AI drawing & tracing for kids.
Contact: panayiotis.savva@silicondali.com
What this app does
Doodle Magic AI lets a child type a kid-safe prompt — "a dragon flying over a castle", "a friendly dinosaur in space" — and an on-device AI engine generates a sketch they can then trace onto real paper through the camera.
Everything runs on the device. No prompts, images, or activity ever leave the phone or tablet.
How to use it
- First launch. The app downloads a small (~72 MB) on-device kid-safe content classifier from this HuggingFace org. This is one-time and required before any generation.
- Pick an AI engine. Tap Create with AI → Download on a model. Engine sizes range from ~1 GB (SD 1.5 Turbo, faster) up to ~2 GB (Absolute Reality, slower but more detailed). You only need one.
- Generate. Type a prompt, pick a Style (Sketch, Anime, Cartoon, Realistic, Watercolor, Pixel Art, Line Art, Low Poly, Fantasy Art), tap Generate. Generation takes 5 seconds to a couple of minutes depending on the model and the device.
- Trace. From the result screen tap View in Gallery, then Trace. Hold the phone over a sheet of paper — the camera overlays the drawing so you can trace it line-by-line. Slide-to-unlock keeps accidental taps from disrupting the trace.
Common questions
Does it need internet? Only for one-time downloads (the safety classifier and the AI engine of your choice). After that, full offline use forever.
Is any data sent anywhere?
No. Prompts and generated images stay on your device. The only network traffic is fetching the safety bundle from huggingface.co/silicondali/doodle-magic-safety and (optionally) AI engines from huggingface.co/silicondali/doodle-magic-models. We don't run any analytics, telemetry, or crash reporting.
Why does it ask for camera and photo permission?
- Camera — to overlay the AI image on a sheet of paper for tracing. Frames are never saved or sent.
- Photo Library — only used when you tap Import Image to trace one of your own photos.
Generation is slow on my phone. SD 1.5 inference is computationally heavy. Older devices take longer. SD 1.5 Turbo is the fastest option and is recommended for older hardware.
A model won't download / download stuck. Check your network. The downloads are resumable — kill and reopen the app, then tap Download again.
The app crashed when I tapped Generate. This was an issue in v4.3.0 on Android. v4.3.2 fixes it (XNNPACK weight-cache versioning) and existing installs heal themselves on first launch — no manual action required.
How do I delete a model? Settings → Manage Models → swipe left on a model → Delete. Frees ~1–2 GB.
Is the AI generation kid-safe? The app runs a pre-generation prompt classifier and a post-generation image classifier (DiffGuard + Marqo, both on-device). Anything that fails either check is blocked. The classifiers are state-of-the-art for the size envelope but not perfect — supervision is still recommended.
My kid drew something the AI didn't generate well — can it learn? No. The on-device model is fixed and never trains on your device's data. Try rephrasing the prompt or picking a different style.
Reporting a problem
Email panayiotis.savva@silicondali.com with:
- Your device model + OS version (e.g. iPhone 15 Pro, iOS 18.4)
- App version (Settings → About → Version)
- What you were doing when the issue happened
- Whether it's reproducible
Reports help us improve. Privacy stays intact — we never need to see your prompts or generated images.
Privacy & terms
- Privacy Policy
- Terms of Use are shown in-app on first launch and acknowledged with the I understand button.
- Downloads last month
- 37