Day page

20 Nov 2025

Paper summary

Paper summary: Kandinsky 5.0: A Family of Foundation Models for Image and Video Generation

Kandinsky 5.0 is an open family of image and 10-second video generation models (6B Image Lite, 2B Video Lite, 19B Video Pro), with the full data-curation and multi-stage training pipeline described.

Three line-ups trade speed for quality: a 2B fast video model up to a 19B high-quality one.
The authors trace their weaker prompt following to the text encoder's 256-token context limit.

Read the post →

AI updates

Nano Banana Pro (Gemini 3 Pro Image) for developers

Google DeepMind released Nano Banana Pro, an image generation and editing model built on Gemini 3 Pro, with more accurate text rendering, Google Search grounding, fine camera and lighting control, and up to 4K output.

It's built on Gemini 3 Pro, so it can use world knowledge, and even Google Search grounding, while generating images.
Developers get explicit control over lighting, camera, focus and colour grading, at 2K or 4K.

Read the post →