Show HN: SceneSynth – Text → scene graph → AI render, recursively (github.com)

SceneSynth is an innovative desktop application that transforms natural language descriptions into detailed, interactive scene graphs, powered by Google’s Gemini and a multimodal image generation model known as Nano Banana. By entering a prompt like "A medieval fantasy kingdom with diverse regions," users can generate complex, explorable world structures that can be recursively expanded from continents to individual rooms. The application allows for meaningful node placement and hierarchical world-building, providing a unique way to visualize and modify scenes through natural language commands. This project is significant for the AI/ML community as it merges natural language processing with visual generation, enabling creators to intuitively design game levels or narrative environments. The ability to modify scene graphs via chat, along with the versatility in rendering styles—ranging from photorealistic to pixel art—highlights its potential applications in game design, storytelling, and virtual reality. The integration of Google Cloud services is essential for its functionality, requiring a Google Cloud account and API setup to access the underlying AI capabilities. SceneSynth not only fosters creativity among developers and artists but also showcases the growing convergence of text-based AI and visual design technologies.