Sogni: Learn logo

Safe Content Filter

Sogni Studio includes an on-device NSFW classifier that screens generated images before they're shown in the gallery. When the filter is on, results flagged as not-safe-for-work are blurred and tagged so you can choose whether to view, delete, or keep them.

#Where to find it

In the right-hand Manual Controls sidebar, expand the Advanced section. The Safe Content Filter toggle is there alongside Sampler, Processing, and Preview Count.

  • Safe Content Filter ON/OFF — turn the classifier on or off.

In Video Create mode the same toggle lives in the Render Settings panel.

#How it works

The filter is a small Core ML model that runs locally on your Mac via the Vision framework — every generated image is classified after it lands and before the gallery thumbnail renders. Nothing about your generations leaves the device for filtering purposes; the classifier weights live inside the app bundle.

When an image is flagged:

  • Its gallery thumbnail is blurred.
  • A tap / click reveals the unblurred version with a one-time confirmation.
  • The image is tagged in metadata so you can filter the gallery by NSFW status later.

#When to use it

  • Public demos / streaming. Keep it on so an unexpected result doesn't make it onto a shared screen.
  • Shared family Macs. On at startup is the right setting.
  • Concept art / private work. Off is fine when you're the only one looking at the gallery.

The filter is a screen for content visibility, not a hard generation block — the underlying model still produced the image, and clicking through always works. If you need a hard block at generation time, choose models tuned for safe content and keep prompts on the safe side.

#See also