Hy Image3.5 preview is now available in ComfyUI. Professional-grade image generation, +30% win rate in human eval vs Hy Image3.0 → Text to image and Image to image in one model, up to 2K → Multilingual text, symbols, and small print that render correctly → Cinematic, comic, commercial photography,... and illustration styles → Identity and product features that hold through scene, outfit, and style changesshow more

Api or open source?

Try the workflow:

That 30% win rate in human eval might just reflect a preference for generic aesthetics rather than the niche, rigorous constraints of actual production workflows.

keep building team!

Hy Image 3.5 in ComfyUI is pro-level now sharper, smarter, and 30% more loved than 3.0.

Which scripts did the small-print test cover? Latin is the easy case. A tiny Thai or Devanagari sign in the background is where I'd look first.

30 percent win rate jump and 2K multilingual text rendering is huge

That braided river texture alone shows this handles fine detail way better than most generators

Pro grade 2K generation with perfect text and identity hold Hy 3.5 is a huge leap

Hy Image3.5 brings impressive control and consistency to ComfyUI

The consistency across styles, details, and identity is seriously impressive. ComfyUI keeps raising the bar.

Huge upgrade 2K, multilingual text that actually works, and identity that holds across scenes. Pro-level stuff!

Check DM

The improved text rendering and identity consistency are the features that really stand out here. 2K generation plus strong style flexibility makes this a compelling upgrade.

Hy Image3.5把文字、图生图和小字渲染补齐了，预览阶段看着很香。但做视频的人最后关心的不是单张赢评测，而是同一角色跨镜头别变脸、细节别闪。静帧王者一进时间轴，才知道谁在裸奔。