Tencent Releases Hy Image3.5 Preview for Iterative Creation
Tencent’s Hy Image3.5 preview adds multi-turn image editing and reference inputs, targeting production workflows rather than one-shot generation.
Tencent has released a preview of Hy Image3.5, its latest image-generation model, with support for text-to-image, image-to-image and multi-turn conversational editing. The model is being introduced through Tencent’s Yuanbao and ima applications, as well as developer-facing access, giving users a way to refine an image across successive instructions instead of restarting from scratch.
From prompts to revision loops
Tencent says an internal blind evaluation involving hundreds of professional designers produced a 30% higher win rate than Hy Image3.0. The company also said the preview performed roughly on par with ByteDance’s Seedream 5.0 Pro and slightly ahead of Qwen-Image-3.0 Pro and Google’s Nano Banana Pro in that internal comparison. Those figures are company-reported and should not be treated as an independent leaderboard result.
The product supports reference-image inputs and conversational context, allowing a creator to alter composition, lighting, subjects or details over multiple turns. Tencent has also promoted high-resolution output and API access, positioning the model for design and content-production workflows rather than casual experimentation alone.
Why it matters
The significance is less about another model entering an already crowded image-generation field than about where Tencent is competing. The company is pairing model quality with direct placement inside a large consumer ecosystem and with a workflow built around repeated edits. That could make the model useful to designers who need controlled variations, not merely attractive first drafts.
The unresolved issue is independent evidence. Tencent’s comparison is based on an internal designer test, and the preview status means reliability, pricing and access limits may still change. Even so, the launch shows Chinese platforms are increasingly treating image generation as an integrated creative tool, with conversational editing and application distribution becoming as important as raw image quality.