cy20251210

We built Seedance2Pro to explore audio-driven video generation.

The tool focuses on turning speech or music into short videos with consistent characters and 2K output, aiming to reduce the time needed for creators to produce visual content.

Happy to hear feedback from the community.

This post explores how AI video stylization has evolved from frame-based style transfer to diffusion-based unified video models, and why temporal consistency became a key challenge.

I wrote this as a technical overview rather than a product announcement. Feedback and discussion are welcome.

GLM-Image is a multimodal image model designed for text-to-image generation and visual understanding.

It supports prompt-based image generation, image editing, and multimodal reasoning, aiming to provide a unified image AI foundation for developers and creators.

This site showcases model capabilities, example outputs, and technical background. It’s especially useful for people exploring AI image generation models or building AI-powered creative products.

Happy to get feedback on the model’s capabilities, real-world use cases, or comparisons with other image models.