On September 22, Tencent's Hunyuan officially released the Hy Image 3.5 preview version, which supports both text-to-image and image-to-image generation, allowing users to upload up to five reference images per session. The model offers multiple aspect ratio outputs and achieves resolutions up to 2K, while also enabling multi-turn conversational creation, letting users continue generating and editing based on historical context.
Hy Image 3.5 preview retains the design philosophy of the Hy Image 3.0 series, delivering notable enhancements in text rendering and layout, realism and stylistic expressiveness, and editing consistency. Priced at a highly competitive rate, it provides exceptional image generation capabilities for both daily creativity and professional visual production, with overall performance markedly improved over its predecessor.
Where to begin
The upgrade in Hy Image 3.5 preview centers on a systematic shift "from intent understanding to visual expression." It is designed to serve everyday scenarios such as photo retouching, portrait beautification, and knowledge cards, while also supporting professional use cases like commercial posters, storyboard creation for series production, film and TV promotional materials, UI design, creative illustrations, brand visuals, and e-commerce assets. The goal is to elevate outputs from merely "looking good" to meeting the demands of "professional-grade aesthetic scenarios."
As a high-value professional image generation model, Hy Image 3.5 preview underwent blind testing by hundreds of internal professional designers, achieving roughly a 30% improvement over Hy Image 3.0. Through deep integration with products such as Yuanbao, WorkRally, OnSolo, Miora, WorkBuddy, and ima, it significantly boosts effectiveness across life, learning, and productivity applications.
Pricing strategy
On the cost front, Hunyuan has further improved inference resource efficiency to underpin cost-effective image generation. For Hy Image 3.5 preview, the Tencent Cloud API pricing for 2K images is set at RMB 0.15 per image, offering strong competitive value. Additionally, charges apply only to generated output images, not to input.
The model is now live across multiple native AI applications, including Tencent's AI assistant Yuanbao, the professional film and TV intelligence platform WorkRally, the specialized series and game content creation tool OnSolo, the creative agent Miora, the office agent workbench WorkBuddy, and the AI knowledge manager ima. API access is concurrently available on Tencent Cloud TokenHub, Tencent Cloud Media Processing Service (MPS), and Cloud VOD.