Z-Image Turbo Free Online
Z-Image Turbo is Alibaba's fast distilled image model — 8 inference steps, results in 2-5 seconds, and surprisingly good detail for the speed. It is the cheapest model in our lineup (5 credits ≈ $0.005 per image) and the default choice for rapid ideation, full-scene rewrites, and bulk generation. Use it free here for text-to-image and image-to-image. 100 free credits on signup.
Aspect Ratio
Advanced Settings▼
Seed for reproducible results (-1 for random)
File format for the generated image
Generated Images
No images generated yet
Configure parameters and click Generate to create images
Z-Image Turbo Specs
What you get when you pick this model — verified through real testing on our platform.
2-5 second generation
Z-Image Turbo runs 8 distilled inference steps. Most generations finish before you can blink. Compared to GPT Image 2 (1-2 min) it's 20-30× faster — perfect for iteration.
Lowest credit cost: 5 credits / image
Equivalent to about $0.005 per generation on our pricing. A $14.90 Basic credit pack buys roughly 1,200 Z-Image Turbo images — the best price-per-image ratio in the lineup.
7 aspect ratios, 720p HD for landscape
Square (1:1) at 1024×1024, landscape (16:9, 4:3, 3:2) and portrait (9:16, 3:4, 2:3) at 1280px on the long edge. Non-square ratios match the 720p HD standard that feels appropriately high-res.
Adjustable strength for image-to-image
Strength parameter (0.0-1.0) controls how much the source image is rewritten. Default 0.7 gives clear changes while keeping composition. Push to 0.85+ for dramatic full-scene rewrites.
JPEG / PNG / WebP output
Three output formats. JPEG and WebP are smaller and faster to download; PNG is lossless for further editing. Z-Image Turbo's per-image cost is the same regardless of format.
Built-in safety checker
fal's content safety filter runs on every generation. Policy-violating outputs are blocked and credits auto-refunded — you never pay for content that doesn't reach you.
What Z-Image Turbo Excels At
Where this model is the clear right choice — based on real testing against the other three models in our lineup.
Rapid ideation and brainstorming
When you want to test 20 variations of a concept in five minutes, only Z-Image Turbo can keep up. 2-5 second cycle time means you iterate at the speed of thought, not at the speed of OpenAI's servers.
Social media content at scale
Instagram posts, TikTok thumbnails, Twitter/X cards. Z-Image Turbo at 1:1 or 9:16 is fast enough that you can generate a week of content in one sitting, picking the best 7 of 50.
Concept art and mood boards
During exploration phase, you don't need polished output — you need lots of options. Z-Image Turbo at strength 0.8+ in image-to-image mode is unbeatable for atmospheric rewrites that explore directions quickly.
Illustrations and flat design
Vector-style, flat color, isometric, low-poly renders all work well. Z-Image Turbo handles these styles competently and you can keep iterating on the same seed to refine.
Bulk content for blog and SEO
Every blog post needs a hero image. At $0.005 each, generating 100 blog featured images costs $0.50. The economics make it feasible to add custom imagery to every post instead of relying on stock photos.
Z-Image Turbo vs. The Lineup
Quick comparison — when to switch from Z-Image Turbo to one of the other three models on the platform.
Switch to GPT Image 2 when you need polish
Final marketing assets, hero images for a campaign, anything that needs text rendered accurately. GPT Image 2 produces ~10× better quality, but at 30× the time and 40× the cost. Use Z-Image Turbo to explore, GPT Image 2 for the final.
Learn more →Switch to Nano Banana 2 for surgical edits
Z-Image Turbo image-to-image rewrites the whole scene. If you need 'change just the jacket color while keeping the face identical', Z-Image Turbo will fail — its diffusion model can't do localized changes. Nano Banana 2 was built for this.
Learn more →Switch to Seedream 4.5 for photorealism
Z-Image Turbo's photorealistic output is decent but not class-leading. For product photography, food shots, or portraits where realism matters, Seedream 4.5 produces noticeably more lifelike skin tones, lighting, and material rendering.
Learn more →Examples
Z-Image Turbo Output Gallery
Real prompts and outputs from Z-Image Turbo. Notice how well it handles stylized illustrations and flat design — that's a sweet spot for this model.
Low-poly character
Low-poly fox standing in a stylized forest, geometric facets, soft pastel color palette, isometric perspective
Flat city skyline
Minimalist flat illustration of a city skyline at sunset, gradient sky, layered silhouettes, vector style
Isometric workspace
Isometric illustration of a cozy home office, plants, books, laptop with steam from a coffee cup, warm afternoon light
Robot companion
Friendly cartoon robot companion, large expressive eyes, soft pastel color scheme, simple background
Brand mascot
Friendly brand mascot character holding a sign, flat vector style, vibrant colors, simple geometric shapes
Story background
Painterly fantasy landscape, distant mountains, rolling meadows, soft volumetric lighting, illustration book style
Getting the Most Out of Z-Image Turbo
Five habits we've learned from testing thousands of Z-Image Turbo generations.
1. Treat speed as a feature
Don't try to nail the perfect prompt on the first generation. Generate 5 quick variants with slightly different wording, pick the best one, then refine. Total time: 15 seconds. With slower models you'd still be waiting for the first result.
2. Lock the seed for refinement, random for exploration
Exploring concepts? Random seed — you want maximum variety. Refining a result? Lock the seed (Advanced Settings) and only change the prompt — same base composition, different details.
3. For image-to-image, start at strength 0.7
0.7 is our default and the sweet spot for clear changes while preserving composition. If the image barely changes, push to 0.85. If too much changes, drop to 0.5. Adjust in steps of 0.05-0.1.
4. Add style and lighting keywords
'Flat vector style', 'isometric clay render', 'soft pastel palette', 'golden hour light' — these short phrases dramatically change the output. Z-Image Turbo is responsive to style cues despite being a fast model.
5. Don't try to render text in the image
Z-Image Turbo struggles with rendering readable text inside images. If your prompt needs accurate text (a poster headline, product label), switch to GPT Image 2 (Medium or High) or Seedream 4.5 — they handle in-image text far better.
Z-Image Turbo FAQ
Try Z-Image Turbo now
100 free credits on signup. No credit card required. First image in 2-5 seconds.
Disclaimer: Image-AI.run is an independent product and is not affiliated with, endorsed by, or sponsored by Alibaba. Z-Image Turbo is developed and owned by Alibaba. All model names, logos, and trademarks are the property of their respective owners. Image-AI.run accesses Z-Image Turbo through the fal.ai platform.