AI-based educational avatars are transforming learning and teaching by offering personalized, contextualized, immersive, and scaffolded instruction. These avatars, powered by large language models (LLMs) and generative AI, can interact with users in natural language, provide feedback, and adapt to individual learning needs. They can be used in various educational settings, including language learning, science education, and tutoring for individual learning difficulties. AI-based avatars can also support contextualized instruction by simulating real-world scenarios and enhancing collaborative learning. Additionally, they can foster self-regulation, interest, and affect by providing personalized feedback and engaging learners through interactive and emotionally responsive interactions. However, the use of AI-based avatars also presents challenges, including the potential for incorrect and inaccurate information, inadequate human relationships, inappropriate values, and insufficient data protection. These challenges highlight the need for careful implementation and ethical considerations in the use of AI-based avatars in education. The article also discusses the importance of developing open-source tools like GPTAvatar, which can be used to create custom AI-based avatars for various educational purposes. The authors conclude that further research is needed to address open questions regarding the effective use of AI-based avatars in education, including how to maximize their benefits while minimizing their risks. The study emphasizes the importance of developing ethical guidelines and regulatory frameworks to ensure the responsible use of AI-based avatars in educational settings.AI-based educational avatars are transforming learning and teaching by offering personalized, contextualized, immersive, and scaffolded instruction. These avatars, powered by large language models (LLMs) and generative AI, can interact with users in natural language, provide feedback, and adapt to individual learning needs. They can be used in various educational settings, including language learning, science education, and tutoring for individual learning difficulties. AI-based avatars can also support contextualized instruction by simulating real-world scenarios and enhancing collaborative learning. Additionally, they can foster self-regulation, interest, and affect by providing personalized feedback and engaging learners through interactive and emotionally responsive interactions. However, the use of AI-based avatars also presents challenges, including the potential for incorrect and inaccurate information, inadequate human relationships, inappropriate values, and insufficient data protection. These challenges highlight the need for careful implementation and ethical considerations in the use of AI-based avatars in education. The article also discusses the importance of developing open-source tools like GPTAvatar, which can be used to create custom AI-based avatars for various educational purposes. The authors conclude that further research is needed to address open questions regarding the effective use of AI-based avatars in education, including how to maximize their benefits while minimizing their risks. The study emphasizes the importance of developing ethical guidelines and regulatory frameworks to ensure the responsible use of AI-based avatars in educational settings.