Take it browsing.

The companion extension runs the hotdog model on your device. It does not use the publisher's API key or upload images. Model files are included in the download.

Download the Chrome / Edge extension

1. Download and extract the ZIP into a permanent folder.
2. Open your browser's Extensions page. Enable Developer mode, choose Load unpacked, and select the extracted folder containing manifest.json.
3. Open a regular webpage, click the not hotdog extension icon, and choose Enable on this tab.
4. Hover over a fully visible image for a moment. The first check loads the local model. Press Escape to pause. Navigation resets activation.

Chrome and Edge Manifest V3 are targeted. Installing the OpenAI plugin does not install this companion. It is distributed as an unpacked preview and is not listed in a browser extension store.

Visible top level HTML images are supported. CSS backgrounds, frames, video, and protected browser pages are outside this preview. When canvas access is blocked, a local visible tab capture is cropped to the image. Overlays may affect the crop. This ImageNet model works best with clear photographs. Each completed check returns HOTDOG or NOT HOTDOG. Drawings and unusual presentations can be misclassified. The model download is about 17 MB and the first check can take a moment.

For an image shared inside ChatGPT or Codex, use the separate not hotdog plugin. That workflow uses the host assistant and its normal plan limits.