InfoVis: Unified Slide Image Generation and Editing

One 9B model for text-to-slide generation, template-conditioned generation, multi-image grounded generation, and instruction-based slide editing.

📄 Paper | 🌐 Project page | 💻 Code | 🤗 Model | 📊 VisSlideBench | 📦 Training data

Terms of use. By using this demo, you agree to use it only for non-commercial research and to follow the FLUX Non-Commercial License. Do not use it to create sexual content, content involving minors, non-consensual or deceptive images of real people, or any violent, hateful, harassing, or otherwise unlawful content. All outputs are AI-generated, and you are responsible for the content you create and how you use it.
Task
Aspect ratio
10 50
1 8
Examples (good cases from the paper)

Tips. InfoVis follows detailed instructions: describe the title, the text, the layout, colors, and visual elements. Text in quotes is rendered as written. For template-conditioned generation, the reference slide sets the style; for multi-image grounded generation, the uploaded images are placed on the slide; for editing, describe the change to make.

Settings. Output size follows the evaluation setup of the paper: 1280x720 for text-to-slide, 1376x768 for template-conditioned and multi-image grounded generation, and the source size (up to about 1 MP) for editing. The demo defaults to 30 sampling steps to save GPU time; the paper uses 50.

License. InfoVis is a derivative of FLUX.2-klein-base-9B and is released under the FLUX Non-Commercial License. This demo is for non-commercial research use.