Elisabeth L. 
The guides so far have covered censoring styles, detection zones, and the various modes. This one covers everything else — the general settings that control how the extension processes content, and the performance options that let you tune it for your specific hardware.
We strongly recommend installing an ad blocker alongside DeniedPixels
Ad-heavy sites load dozens of extra images and autoplay videos (we’ve seen up to 3-4 videos censoring processes at the same time!) that the extension has to process on top of the actual content. On mobile hardware, this can cause noticeable lag. An ad blocker removes the junk before DeniedPixels ever sees it, resulting in smoother video censoring and better battery life. Check out our ad blocker performance guide for more details.
General settings
Loading Placeholder controls what you see while the AI is still processing an image. When enabled, images show a blurred placeholder until detection is complete. When disabled, the original image is visible during the brief processing window. If you want zero chance of seeing unprocessed content, keep this on.
Minimum Image Size tells the extension to skip images smaller than a set pixel size. Default is 50px. Tiny thumbnails, icons, and UI elements don’t need to be scanned — skipping them saves processing time without affecting your actual experience. Raise it if you want to ignore more small images, lower it if you want everything scanned.
GIF Censoring enables detection on animated GIFs. This is more demanding than static images since the extension needs to analyze multiple frames. If you don’t care about GIF content or your machine is struggling, you can turn it off.
Embedded Video Censoring handles video embeds from known adult sites — for example, RedGIFs links embedded in Reddit posts. When enabled, these embeds get blurred automatically. This is a lightweight feature since it works off domain detection rather than AI analysis.
Video Censoring is the big one. This enables AI-powered detection on video content. It has three modes:
- AI Detection runs the full detection pipeline on video frames, censoring specific zones just like it does with images. This gives you the most precise results but is the most hardware-intensive option. Performance depends heavily on your GPU and CPU.
- Full Blur takes a lighter approach — it blurs the entire video player on known adult sites. No per-zone detection, just a blanket blur. Much easier on your hardware.
- Black Box is similar to Full Blur but covers 90% of the video player with a black overlay on known adult sites. Also lightweight since it skips AI detection entirely.
If your machine handles AI Detection smoothly, use it — you get the same granular zone-based censoring on video that you get on images. If you’re seeing lag or frame drops, Full Blur and Black Box give you coverage without the processing cost.
Performance: Images
Max Concurrent Images sets how many images the extension processes at the same time. Default is 3. On a fast machine you can raise this so pages with lots of images get scanned faster. On a slower machine, lowering it prevents the extension from hogging too many resources at once.
GIF Hover-Only Mode is a useful optimization. When enabled, GIFs and animated WebP images only get loaded and censored when you hover over them. Everything else stays blurred by default. This saves significant CPU and memory on pages with lots of animated content — the extension only does the work when you’re actually looking at something.
GIF Detection Interval controls how many frames the extension analyzes in animated images. A value of 1 means every frame gets scanned — smoothest censoring, heaviest load. The default of 5 means every 5th frame, which is a good balance. Values of 10 or higher are lighter on CPU but might miss brief scene changes in longer animations.
Performance: Videos
Video performance settings give you fine-grained control over how often the AI samples frames during playback. These only apply when Video Censoring is set to AI Detection.
Detection Frequency controls how often the AI samples video frames during playback. It’s a slider from 1 to 10. At 1 (battery saver), the extension checks every 1.5 seconds — lightest on your device but censoring may lag behind fast scene changes. At 10 (fastest), it checks every 0.1 seconds for near-real-time tracking — GPU recommended at this level. If video playback stutters, bring the number down. If censoring feels delayed, bring it up. Default works for most setups, but if you have a dedicated GPU, cranking it to 10 is worth trying.
Detection Persistence controls how long censor overlays stay visible when the AI briefly loses detection between frames. Video detection isn’t perfect frame-to-frame — the model might detect content on one frame, miss it on the next, then detect it again. Without persistence, you’d see the censor overlay flash on and off. The default of 10 means overlays stay visible for 10 missed frames before disappearing. Raise it to 20+ if you’re seeing flashing. Set it to 1 if you want overlays to disappear immediately when detection drops.
AI Backend
Inference Backend determines how the AI model runs on your machine. Three options:
- Auto is the recommended default. The extension tries to use your GPU first and falls back to CPU if that doesn’t work. This gives you the best performance your hardware can handle without any manual configuration.
- GPU Only forces all AI processing onto your graphics card. Faster than CPU but won’t work if your GPU doesn’t support the required features or drivers.
- CPU Only forces everything onto your processor. Slower but universally compatible. Use this if you’re experiencing crashes or visual glitches with GPU mode, or if you want to keep GPU usage low for other applications.
Changing the inference backend requires an extension reload to take effect. If you’re not sure which to use, leave it on Auto.
