We’re launching Lyria 3.5 in Google Flow Music, with advances across musicality, lyrics, vocals, and creative control
deepmind.google·6d ago
AI Engineering
Hugging Face Blog·Jun 9·5 min read
How an Agent Built a 3D Paris Gallery by Chaining Two Hugging Face Spaces
TL;DR
Integrating various multimedia AI models has traditionally been complex due to differing formats and requirements. A coding agent utilized two Hugging Face Spaces to create a 3D gallery of Parisian monuments without manual intervention.
✦ Why It Matters
Engineers can leverage modular AI components to streamline multimedia application development and reduce integration complexity.
Key Takeaways
- Utilize Hugging Face Spaces for rapid multimedia development.
- Leverage agents to automate integration of AI models.
- Experiment with chaining outputs from different Spaces.
- Reduce development time by reusing existing components.
How It Works
The agent first generated images of Paris monuments using the ideogram-ai/ideogram4 Space, which produced clean images suitable for 3D reconstruction. It then fed these images into the VAST-AI/TripoSplat Space to create 3D Gaussian splats.
The agent handled integration tasks, such as adjusting orientations and compressing files, and built a Three.js viewer for interaction.