DialogLab is a new open-source framework designed to facilitate dynamic human-AI group conversations. Developed by Erzhen Hu and Ruofei Du at Google XR, it addresses the limitations of one-on-one interactions with AI by allowing for more complex, multi-party dialogues. This tool lets creators define agent personas, manage group dynamics, and specify turn-taking rules, offering a blend of structured scripting and improvisation that reflects real-life conversations. The framework operates through a streamlined workflow: author, test, and verify. Users can create conversational scenarios using a drag-and-drop interface, which simplifies the scene setup. During testing, a live preview panel shows the ongoing conversation, while a "human control" mode enables designers to tweak AI responses. Participants in user evaluations found the human control mode to be significantly more engaging and realistic compared to fully autonomous or reactive agents. They appreciated the interface's intuitiveness and the balance between auto-generated prompts and manual adjustments. Feedback from 14 participants across various fields highlighted DialogLab's effectiveness for simulating real-world conversations, especially in educational settings or social science research. Users noted its ability to model different dialogue strategies and the utility of the verification dashboard for analyzing conversation dynamics. Future enhancements may include integrating non-verbal cues and photorealistic avatars for more immersive simulations, expanding DialogLab's applications beyond its current capabilities.