Tencent has launched the image-to-image version of its Hunyuan Image 3.0 model, expanding its generative AI capabilities in image editing and multi-image blending, the company announced.

The newly released model, known as HunyuanImage 3.0-Instruct, supports a wide range of image editing functions, including object addition and removal, style transformation, image enhancement, and multi-image blending. The model is built on an 80-billion-parameter mixture-of-experts architecture and has been trained through instruction fine-tuning and post-training to achieve a deep understanding of both user-provided images and editing commands.

Tencent said the model adopts a “think first, then edit” workflow. After receiving an input image, the system analyzes visual content and reasons through the required editing steps before generating output, helping ensure consistency and accuracy in the final results.