AI Documentation

PV VisionGen uses Artificial Intelligence to generate photorealistic visualizations of photovoltaic systems on buildings and open spaces. The AI analyzes uploaded photos and generates modified versions showing PV modules on the respective location.

The system is classified as LOW RISK under the EU AI Act. PV visualization is not a high-risk application. The system does not make autonomous decisions affecting people's rights or safety.

- Google Gemini (image generation/inpainting): Used to generate photorealistic PV module placement on uploaded photos. - Anthropic Claude (image analysis & validation): Used to analyze uploaded photos (perspective, terrain, lighting conditions) and validate generated results for plausibility. - FLUX.1 Fill [dev] (image inpainting): Alternative model for PV module inpainting.

- Images are sent to Google and Anthropic APIs for processing. - No training: Uploaded images are NOT used to train AI models. Both Google and Anthropic offer API terms that exclude training. - Images are deleted within 24 hours from our servers. - Processing is based on Art. 6 (1) lit. b GDPR (contract fulfillment).

- Results are AI-generated approximations, not engineering plans. - Module placement may not reflect actual structural feasibility. - Lighting and shadow conditions are approximated. - Results vary depending on image quality and perspective. - The system cannot assess roof load capacity or electrical specifications.

Users review all generated results before use. No automated decisions are made. The system is a visualization tool, not a planning or engineering tool.

For questions about AI usage, please contact us at: info@pv-visiongen.de