Z-Image
Z-Image Turbo from Alibaba's Tongyi-MAI lab: a small, open 6B model that makes realistic pictures in 8 steps, with English and Chinese text, at the lowest price in the Lab.
What it nails
- The cheapest pictures in the Lab
- Realistic portraits in 8 steps
- English and Chinese text
Watch out for
- Text only: no reference pictures
Z-Image is the open picture model from Tongyi-MAI, a lab at Alibaba. The version in the Lab, Z-Image Turbo, is small (6 billion parameters, free under the Apache-2.0 licence) and built for speed: it needs only 8 steps, and its makers report under a second per picture on their GPUs. Despite its size it is strong at realistic portraits and products, and it writes English and Chinese text accurately.
It is also the cheapest picture in the Lab, so it is the best place to try ideas: run 4 at once and keep the best.
At a glance: Z-Image. The live spec card, price calculator and Generate button show here on the site.
What Z-Image is secretly great at
Realistic people, for pennies
Tongyi-MAI highlights "highly realistic portraits and product images". Name the camera and the film and describe the skin, and Turbo keeps the texture.
A hyper-realistic, close-up portrait of a tribal elder from the Omo Valley, painted with intricate white chalk patterns and adorned with a headdress made of dried flowers, seed pods, and rusted bottle caps. The focus is razor-sharp on the texture of the skin, showing every pore, wrinkle, and scar that tells a story of survival. The background is a blurred, smoky hut interior, with the warm glow of a cooking fire reflecting in the subject's dark, soulful eyes. Shot on a Leica M6 with Kodak Portra 400 film grain aesthetic.
Example by fal.ai · source ↗
Text in English and Chinese
Put the words in quotes and say where they are. Z-Image renders signs, posters and panels that read correctly, in both languages.
Photo of a stylish young woman with short black hair standing confidently in front of a vibrant cartoon-style mural wall. She wears an all-black outfit: a puffed bomber jacket with a ruffled collar, cargo shorts, fishnet tights, and chunky black Doc Martens, with a gold chain dangling from her waist. The background features four colorful comic-style panels: one reads "GRAND STAGE" and includes sneakers and a Gatorade bottle; another displays green Nike sneakers and a slice of pizza; the third reads "HARAJUKU st" with floating shoes; and the fourth shows a blue mouse riding a skateboard with the text "Takeshita WELCOME." Dominant bright colors include yellow, teal, orange, pink, and green. Speech bubbles, halftone patterns, and playful characters enhance the urban street-art aesthetic. Daylight evenly illuminates the scene, and the ground beneath her feet is white tiled pavement. Full-body portrait, centered composition, slightly tilted stance, direct eye contact with the camera. High detail, sharp focus, dynamic framing.
Example by Alibaba Cloud · source ↗
Detailed illustration, not just photos
Give it a full art-direction brief (style, palette, composition, mood) and it follows it like a much bigger model.
Finished square editorial illustration in a sophisticated contemporary anime and manga style for a renewable-energy trade publication. An underwater maintenance engineer inspects a full-scale tidal power turbine anchored to the rocky seabed in a cold northern sea. Compose the diver prominently in the lower-left foreground, wearing a practical orange dry suit, matte white commercial-diving helmet, black buoyancy gear, tool harness, pressure gauges, hoses, and work gloves. The diver directs a narrow inspection lamp toward a clearly visible crack sensor and bolted access panel on one turbine blade; one hand steadies the lamp while the other holds an ultrasonic testing probe against the metal. The enormous three-bladed turbine rises diagonally through the center and upper-right of the image, communicating realistic industrial scale. Include believable engineering details: marine-growth patches, rivets, cable conduits, sacrificial anodes, hydraulic joints, a seabed power cable, and a small tethered inspection drone hovering nearby. Suspended particles and streams of bubbles reveal the current; distant fish silhouettes and shafts of filtered surface light add depth without turning the scene fantastical. Art direction: crisp expressive ink contours, disciplined mechanical draftsmanship, restrained cel shading, subtle halftone texture, and selective dry-brush marks inspired by premium seinen manga editorial art. Palette of deep teal, oxidized copper, graphite, muted kelp green, and safety orange, with the inspection beam casting a warm pale-yellow focal glow. Mood: focused, competent, hazardous but controlled. Dynamic low-angle perspective, strong silhouette hierarchy, clean readable anatomy, accurate diving equipment, detailed machinery, no typography, no logos, no interface elements, no border, standalone publication-ready artwork.
Example by Runware · source ↗
Short prompts work too
One sentence is enough for a fun, clean result. Because each picture costs so little, it is the ideal model for quick tests before you move a prompt to a bigger model.
A T-Rex relaxing on a beach, lying on a sun lounger and wearing sunglasses.
Example by AI/ML API · source ↗
Z-Image Turbo in the Lab
| Setting | What you get |
|---|---|
| Pictures per run | Up to 4 |
| Shapes | Square, 4:3, 3:4, 16:9, 9:16 |
| Reference pictures | Not supported: Z-Image Turbo is text-to-picture |
Z-Image is the cheapest picture in the Lab. Test your idea here with 4 pictures, then move the winning prompt to Nano Banana, GPT Image or Seedream for the final version.
How to prompt Z-Image
From the examples Tongyi-MAI and Alibaba Cloud publish.
- Start with the kind of picture: "Photo of…", "Editorial illustration…", "Close-up portrait…".
- Describe it in full: subject, clothes, background, light, composition.
- Quote any text and say where it sits.
- Name a camera or film ("Leica M6", "Kodak Portra 400") for a real-photo look.
- Write in English or Chinese.
Z-Image can think, too
Tongyi-MAI pairs Z-Image with a Prompt Enhancer that reasons and uses world knowledge before drawing: it can illustrate a famous poem, plan a trip or solve a maths problem in a picture. These three official examples were made that way.
In English: "Illustrate the two most famous lines of the poem Deng Ke Hou."
帮我给《登科后》配图,最出名的两句
Example by Tongyi-MAI · source ↗
In English: "Plan a West Lake trip in Hangzhou for me, as a hand-drawn journal page."
帮我规划一个杭州西湖的旅游计划,手帐
Example by Tongyi-MAI · source ↗
Given that chickens and rabbits are in the same cage, there are a total of 35 heads and 94 feet. Find the number of chickens and rabbits.
Example by Tongyi-MAI · source ↗
The Lab sends your prompt to Z-Image exactly as you write it, without the Prompt Enhancer. For results like these, write out what the picture should show (the lines of the poem, the stops of the trip, the answer), not only the question.
FAQ
Why is Z-Image so cheap?
It is a small, fast model: 6 billion parameters and only 8 steps per picture, so each picture costs very little to make.
Is Z-Image good enough for final work?
For realistic portraits, products and illustration, often yes. For long text or complex layouts, use a larger model like Qwen Image or GPT Image.
Can Z-Image write text?
Yes, in English and Chinese. Put the words in quotes in your prompt.
Can I upload a reference picture?
No. Z-Image Turbo makes pictures from text only. For reference-based work, use Nano Banana, Seedream or Kling Image.