A Yolov8 detection model that detects comic book speech bubbles and sound effects in images.

The model can be used as an ADetailer model (for Automatic1111 / Stable Diffusion use), or using other inference scripts to return detection bounding boxes of watermarks.

A small tutorial on how to use the model can be found on this Github: https://github.com/MNeMoNiCuZ/yolov8-scripts or this CivitAI article.

This model is only meant for research purposes. The model is entirely trained on the following dataset: yolomanga/speechballoon_comic However, since the dataset is created entirely out of Marvel comic book panels, I think the original author cannot licence the images as CC4. I do not think this model can ba used commercially either.

comic_speechbubble_m_yolov8_v1:

comic_speechbubble_s_yolov8_v1

Note:

The large preview images may not be from the correct model.

The A1111 screenshots are from the correct version however.

The medium model performs slightly better in general.