I built a minimal web UI to let people try out the recently released Z-Image-Turbo model by the Tongyi-MAI team.
The model is fascinating because it uses progressive distillation to achieve inference in just 8 steps. It’s a 6B parameter model that strikes a really interesting balance between speed (sub-second generation) and photorealism, distinct from the heavier Flux or SDXL models.
Key features of the site:
Fast: Optimized specifically for the Turbo version.
No Login Required: Just open and generate. I wanted to lower the friction for testing the model.
Bilingual: The model handles Chinese prompts natively alongside English.
It’s currently an MVP running on Next.js. I’m covering the GPU costs for now so everyone can try it out freely.
Would love to hear your thoughts on how the image quality compares to Flux/Midjourney in your tests!
I built a minimal web UI to let people try out the recently released Z-Image-Turbo model by the Tongyi-MAI team.
The model is fascinating because it uses progressive distillation to achieve inference in just 8 steps. It’s a 6B parameter model that strikes a really interesting balance between speed (sub-second generation) and photorealism, distinct from the heavier Flux or SDXL models.
Key features of the site: Fast: Optimized specifically for the Turbo version. No Login Required: Just open and generate. I wanted to lower the friction for testing the model. Bilingual: The model handles Chinese prompts natively alongside English.
It’s currently an MVP running on Next.js. I’m covering the GPU costs for now so everyone can try it out freely.
Would love to hear your thoughts on how the image quality compares to Flux/Midjourney in your tests!