Tencent's WeVisOne-VL unifies detection, OCR, segmentation and matting

The open Apache 2.0 vision-language model adds to a fast-expanding field of unified multimodal tools.