Tar by ByteDance

Tar by ByteDance is a unified multimodal large language model (MLLM) designed to bridge visual and language understanding and generation within a single AI framework. It introduces the Text-Aligned Tokenizer (TA-Tok), allowing conversion between images and discrete tokens aligned to LLM vocabularies, thus enabling both image-to-text and text-to-image tasks. This tool is ideal for AI researchers, developers, and organizations seeking cross-modal AI solutions that require advanced visual comprehension and generation alongside language capabilities.