Chinese technology company ByteDance has publicly released its newly developed AI model, Bagel, and made its source code available on GitHub and HuggingFace platforms.
Bagel is primarily designed for working with images, allowing users to generate, edit, analyze graphics, and provide detailed explanations of visual content. The model has 7 billion parameters and stands out for its creative capabilities. It can make realistic edits to images while preserving the style and key elements of the original artwork. For example, when editing the famous Mona Lisa painting, the model was able to maintain its artistic features.
One of Bagel’s notable features is its ability to create anime-style illustrations or 3D figures based on photos. Users can also ask the model to “unfold” an object in an image, meaning to visually expand its details while preserving patterns and designs.
By activating the model’s “reasoning mode,” users can receive deeper analysis and clearer explanations of an image. According to the company, Bagel’s capabilities are comparable to OpenAI’s GPT-4o and Google’s Gemini 2.0 models.
Currently, the demo version is free but may sometimes run slowly or experience interruptions. For more stable performance and integration into personal projects, users are encouraged to install the model locally.
This step by ByteDance demonstrates the company’s commitment to contributing to AI technology through open-source tools and highlights its innovative potential in the field.
