Efficient photoreal images with Z-Image
Alibaba Tongyi-MAI’s 6B S3-DiT image model — fast text-to-image in LazyKiwi.
Z-Image (Hugging Face / GitHub: Tongyi-MAI/Z-Image and Z-Image-Turbo) is a 6B-parameter Scalable Single-Stream Diffusion Transformer (S3-DiT) family from Alibaba’s Tongyi-MAI. It targets strong photorealism and bilingual text rendering with efficient inference — Turbo variants emphasize few-step speed under consumer VRAM budgets. LazyKiwi offers Z-Image as a text-to-image model at promotional pricing (2 credits/image) without reference upload in the current workbench profile.
- Provider
- Tongyi-MAI
- Parameters
- ~6B
- Mode
- Text→Image
- Credits
- 2 / image
How it works
Use Z-Image in three steps
Open the LazyKiwi workbench, pick Z-Image, and generate without leaving the page.
1Choose the model
Select Z-Image from the model list so generation uses the matching upstream endpoint and credit tier.
2Write a prompt (and optional edit image)
Describe the subject, style, lighting, and composition. Some models also accept a reference image for guided edits.
3Generate and export
Run a small batch, compare takes on mobile framing, then download the keepers for posts, ads, or further edits.
Why Z-Image
What Z-Image is built for
Z-Image is available in LazyKiwi with workbench controls aligned to public model capabilities and our production credit table.
S3-DiT efficiency
Single-stream diffusion transformer
Z-Image concatenates text, semantic, and VAE tokens in one stream to keep a 6B footprint while aiming for frontier-like photoreal quality — see the Z-Image technical report and HF Turbo card for architecture notes.
- ~6B parameters
- Apache 2.0 open release
- Turbo few-step variants upstream
Photoreal + bilingual text
Stills that hold up in ads
Upstream positioning emphasizes photorealistic generation and English/Chinese text rendering — useful for posters, packaging mockups, and product scenes.
- Photoreal defaults
- Bilingual text rendering
- High-volume friendly pricing
LazyKiwi profile
Text-to-image, promo credits
Workbench id zimage: text-to-image only (no upload), 2 credits per image with a 20-credit list price shown as 90% OFF.
- No reference upload
- 2 credits / image
- Ideal for batch ideation
Vs Qwen Image
Related but separate Alibaba lines
Z-Image (Tongyi-MAI) and Qwen-Image (Qwen team) are different open image lines. Use Z-Image for cheap photoreal drafts; use Qwen Image when you need single-image edit flows in LazyKiwi.
- zimage ≠ qwen-image
- Both Apache-oriented open stacks
- Pick by workflow, not brand alone
Sample directions
Ideas to try with Z-Image
Start from these directions in LazyKiwi, then refine prompts to match your brand or story.
Studio product still
Natural light portrait
Clean poster layout
Lifestyle cafe scene
Soft concept still life
App promo graphic
Specs
Z-Image at a glance
Capabilities exposed through the LazyKiwi workbench (production limits may be narrower than the open-weight ceiling).
- Provider
- Tongyi-MAI (HF Z-Image / Z-Image-Turbo)
- Type
- Text-to-image (S3-DiT ~6B)
- Inputs
- Text prompt only (LazyKiwi)
- LazyKiwi mode
- Text-to-image
- Credits
- 2 per image (promo)
- List price
- 20 credits (strikethrough)
- Upload
- Disabled for Z-Image
- Best for
- Photoreal drafts at scale
When to use it
Z-Image vs other LazyKiwi image models
Z-Image is the cheapest photoreal open option. FLUX Schnell is a fast generalist; Qwen Image adds edit uploads.
| Model | Strength | Best for | Credits |
|---|---|---|---|
| Z-Image | Efficient photoreal T2I | High-volume drafts | 2 / image |
| Qwen Image | T2I + single-image edit | Design & edits | 2 / image |
| FLUX.1 Schnell | Fast general quality | Broad styles | 10 / image |
Use cases
What creators build with Z-Image
Practical workflows where Z-Image is a strong default inside LazyKiwi.
Lifestyle stills
Generate photoreal scenes for ads without a shoot day.
Poster type tests
Probe bilingual title treatments before designer polish.
Catalog ideation
Batch SKU contexts at 2 credits each.
Prompt mining
Explore dozens of looks cheaply before Nano Banana / Seedream finals.
FAQ
Z-Image FAQ
What is Z-Image?
Z-Image is Tongyi-MAI’s 6B open image generation family (HF: Tongyi-MAI/Z-Image and Z-Image-Turbo) using an S3-DiT architecture, known for efficient photoreal generation.
Can I upload a reference image?
Not in the current LazyKiwi Z-Image profile — it is text-to-image only. Use Qwen Image when you need single-image editing.
How much does it cost?
2 credits per image on the promo tier (list price 20 credits, shown as 90% OFF).
Is Z-Image the same as Qwen-Image?
No. They are separate Alibaba-ecosystem open models. LazyKiwi lists them as distinct workbench entries.
Try Z-Image in LazyKiwi
Generate with Z-Image in the same workspace as your other image and video models.
More models
Other models available in the same workbench.
