Tencent Mixture Releases Hybrid Image 3.5 Preview to Meet the Needs of Professional Aesthetic Scenes

The Zhitong Finance App learned that on September 22, Tencent Mixture officially released the Hy Image3.5 preview (Hy Image3.5 preview). This model supports Wensheng maps and cartoons. Up to 5 images can be uploaded at a time as a reference, and supports output in various size ratios. It can achieve up to 2K resolution output. The model supports multiple rounds of dialogue creation, and can be continuously created and edited based on historical context.

According to the introduction, the Hy Image 3.5 preview continues to follow the design ideas of the Hy Image 3.0 series, with significant improvements in text rendering and layout, realism and stylistic expressiveness, editing consistency, etc., to provide highly effective image generation capabilities for everyday creation and professional visual production with a very attractive cost ratio. Compared with the previous generation model, the comprehensive capabilities of the Hy Image 3.5 preview have been significantly improved.

The Hy Image 3.5 preview has been systematically upgraded around “from intention understanding to visual expression”. It can not only serve lifestyle entertainment and learning scenes such as daily retouching, portrait beautification, and knowledge cards, but also support professional-level scenes such as commercial posters, drama segments, film and television materials, UI design, creative illustrations, brand vision, e-commerce materials, etc., so that the generated results move from “good looking” to meet “professional aesthetic scene needs”.

This is a cost-effective professional-grade graphic model. After blind testing by hundreds of in-house professional designers at GSB, the model's ability was significantly increased by about 30% compared to Hy Image 3.0. Through deep collaboration with products such as Yuanbao, WorkRally, OnSolo, Miora, WorkBuddy, and ima, the effects and capabilities for life, learning, and productivity scenarios have been greatly enhanced.

In terms of price, mixed elements further improve the utilization rate of inference resources and provide a foundation for cost-effective image generation. Hy Image3.5 preview Tencent Cloud API 2K images are priced at 0.15 yuan/image, which is very competitive in cost performance; at the same time, Hy Image3.5 preview only charges for output images.

Currently, the model has launched various native AI applications such as Tencent AI assistant Yuanbao, professional film and television intelligence platform WorkRally, professional series and game content creation tool OnSolo, creative intelligence Miora, office smart workbench WorkBuddy, AI knowledge manager ima, etc., with API capabilities simultaneously launched on Tencent Cloud TokenHub, Tencent Cloud Media Processing MPS, and Cloud On Demand VOD.