Z-Image

Tongyi-MAI's Z-Image foundation model with CFG and step control for text-to-image.

What it is good for

Tunable foundation generation with Z-Image

Tongyi-MAI’s Z-Image foundation model from Pruna adds CFG and step control for text-to-image when you need tighter prompt adherence.

Dial guidance for briefs that must match brand or layout constraints.

Generate posters, packaging mocks, and UI comps with readable text.

Experiment with steps and CFG before committing to a look.

Key capabilities

Adjust guidance scale and inference steps for literal or exploratory results.

Higher fidelity than Turbo when the brief needs cleaner detail.

Better on-image typography and label work than the Turbo tier.

Standard framing for social, product, and presentation layouts.

Foundation weights suited to custom style workflows beyond defaults.

How to get a strong result

Start with the subject, then add lighting, style, and anything that must stay the same. Upload a reference whenever you want a more controlled edit.

1
Start near default CFG
Guidance around 4 balances adherence and creativity for most prompts.
2
Increase steps for finals
More steps can sharpen detail at the cost of latency.
3
Draft on Turbo first
Validate direction on Z-Image Turbo, refine winners on Z-Image.
4
Document winning pairs
Save step and guidance settings that repeat well.

Series compare

Z-Image Turbo for sub-second drafts versus Z-Image foundation with CFG — compare Tongyi-MAI text-to-image by speed and control.