Skip to content
CozyGPU

Z-Image Turbo

Fast text-to-image. About five seconds a picture on a warm worker.

Modes
Text to image
GPU class
32 GB to 48 GB Ada or Ampere
RunPod pool price
$1.22 to $1.75/h
Typical cost per run
about $0.003 per image
01

What it is

Z-Image Turbo is a distilled open-weight text-to-image model built for speed: nine steps by default, sharp 1024-pixel output, and strong prompt following for a model this quick. It is the model to reach for when you want to iterate on a prompt dozens of times without watching the meter.

CozyGPU runs the published ComfyUI workflow for you on a mid-size RunPod GPU and ships with a handful of built-in LoRAs (Realistic Snapshot, 35 mm Photo, Pixel Art, DeJPEG, Character Design Sheet). Your own LoRAs drop in beside them.

Weights on Hugging Face: Comfy-Org/z_image_turbo
02

Modes

  • Text to imagePrompt in, one or more images out. Aspect presets from square to 9:16.
03

Settings exposed

  • SizeAspect presets (1:1, 3:2, 16:9, 4:5, 9:16); default 1024 x 1024.
  • Steps5 to 30, default 9.
  • SeedRandom, or fixed to reproduce a result.
  • Sampler and schedulerFull list exposed; default res_multistep / simple.
  • LoRAsBuilt-in style LoRAs plus your own, each with a strength slider and trigger words.
  • BatchSeveral images from one prompt in one job.
  • Custom LoRAsSupported. Built-in LoRAs ship with the worker; your own .safetensors files load by drag and drop.

Other models