I built a small web app for trying out Z Image Turbo, an ultra-fast text-to-image model that generates up to 2K (2048×2048) images and handles English + Chinese text in a single shot: https://zimageturbo.app/
A few details:
– Uses the 6B-parameter Z Image Turbo model from Tongyi-MAI, which fits in 16GB VRAM while still producing high-quality images.
– Supports up to 2048×2048 output and multiple formats (PNG/JPG/WebP).
– Optimized for prompts that mix image description and typography (e.g. posters, covers, logos with both English and Chinese text).
My goal with this site is:
– Make it easy to try the model without setting up GPUs, ComfyUI, or downloading weights.
– Provide a simple, focused UI for fast iteration instead of a “do everything” image lab.
I’d love feedback from HN on:
– Latency and UX (does it feel “instant” enough for you?).
– Prompt handling, especially for bilingual text.
– What knobs / features you’d actually want exposed (seed control, negative prompts, presets, etc.).
JoahYi•43m ago
A few details: – Uses the 6B-parameter Z Image Turbo model from Tongyi-MAI, which fits in 16GB VRAM while still producing high-quality images. – Supports up to 2048×2048 output and multiple formats (PNG/JPG/WebP). – Optimized for prompts that mix image description and typography (e.g. posters, covers, logos with both English and Chinese text).
My goal with this site is: – Make it easy to try the model without setting up GPUs, ComfyUI, or downloading weights. – Provide a simple, focused UI for fast iteration instead of a “do everything” image lab.
I’d love feedback from HN on: – Latency and UX (does it feel “instant” enough for you?). – Prompt handling, especially for bilingual text. – What knobs / features you’d actually want exposed (seed control, negative prompts, presets, etc.).
Link: https://zimageturbo.app/
Happy to answer technical questions in the comments and share more details about how it’s wired up.