Bouncer 2.0 Filters AI-Generated Images and Text From Your X Feed
Imbue's Bouncer lets users filter their X feed. Its 2.0 update added dedicated AI image and text detection; image classification uses a server-side DINOv3 model
Image source: Imbue — Bouncer 2.0 official technical article
Bouncer is an X feed-filtering tool introduced by Imbue in April 2026. It lets people describe content they want removed, using context and tone rather than relying only on keyword matches. In July, Imbue announced Bouncer 2.0, adding dedicated detection for AI-generated images and text.
Separate Models Detect AI Images and Text
In its Bouncer 2.0 technical article, Imbue describes separate detectors for AI-generated images and text. Bouncer's Remove AI Slop control enables filtering of content identified by those detectors.
For images, Imbue uses a fine-tuned 7-billion-parameter DINOv3 model that currently runs on its servers. The image classifier therefore does not run as a small model inside the browser extension.
Text detection follows a different path. Imbue describes a Gemma 4 E2B-based model with a LoRA adapter for AI text detection on iPhones, which operates on the device. The distinction matters: local text detection does not imply local AI image detection.
Custom Feed Filters and Reviewable Results
Beyond AI content detection, users can specify unwanted topics or tones in natural language, including concepts such as arrogance and pessimism. The original Bouncer product description demonstrates manually bouncing a post and accessing a view of filtered posts.
Imbue lists Chrome and Firefox extensions and an iPhone app, while the Bouncer 2.0 announcement also lists Safari. Product information and platform details are available on the Imbue pages below.