An iOS app that uses ChatGPT to generate ARKit code from spoken prompts, placing and manipulating 3D objects in augmented reality.
ChatARKit is an iOS application that uses ChatGPT to generate ARKit code from natural language prompts, enabling users to create augmented reality scenes by speaking commands. It translates spoken instructions into JavaScript that manipulates 3D objects in AR, such as placing models on detected surfaces and animating them. The project is an experiment in leveraging large language models for code generation in custom API environments.
iOS developers and AR enthusiasts interested in exploring AI-driven content creation, particularly those experimenting with ChatGPT integration and natural language interfaces for augmented reality.
Developers choose ChatARKit for its novel approach to AR prototyping, allowing rapid scene creation without manual coding. Its integration of Whisper for speech recognition and Sketchfab for 3D assets provides a streamlined, albeit experimental, pipeline for AI-generated AR experiences.
Using ChatGPT to create AR experiences with natural language.
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Allows creating AR scenes via voice commands, with ChatGPT translating prompts into executable JavaScript for ARKit, enabling rapid prototyping without manual coding.
Uses Whisper.cpp for fast, offline speech-to-text transcription, reducing latency and enhancing privacy compared to cloud-based solutions.
Integrates Sketchfab to fetch models based on object descriptions, with fallback to native cubes, providing flexibility in asset sourcing.
Updated to use direct OpenAI API instead of a web bridge, simplifying setup and improving speed, as noted in the README's usage instructions.
ChatGPT frequently produces non-functional or inconsistent JavaScript, with the README admitting 'highly variable results' and 'bad code that fails to run.'
No on-screen status or error messages; users must rely on Xcode console logs for debugging, making it frustrating for non-developers.
Models often fail to download or import, as highlighted in the README, breaking object placement and requiring manual intervention.