Unified slide image generation and editing: paper, model, VisSlideBench benchmark, and training data.
InfoVis
non-profit
AI & ML interests
Slide generation, image editing, text-to-image
Recent Activity
Organization Card
InfoVis: Unified Slide Image Generation and Editing
One 9B model, fine-tuned from FLUX.2-klein-base-9B, for text-to-slide generation, template-conditioned generation, multi-image grounded generation, and instruction-based slide editing.
| 📄 Paper | arXiv:2610.16784 |
| 🌐 Project page | vectorspacelab.github.io/InfoVis |
| 💻 Code | VectorSpaceLab/InfoVis |
| 🤗 Model | InfoVis/InfoVis |
| 🚀 Demo | Try InfoVis online: generate and edit slides in the browser |
| 📊 Benchmark | InfoVis/VisSlideBench: 1,527 manually screened cases in four tracks |
| 📦 Training data | InfoVis/Data: ≈15M records for continual pretraining and SFT |