Image Editing
Edit instruction
“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
Tongyi-MAI's 6-billion parameter distilled text-to-image model optimized for speed, achieving high-quality generation in 8 steps or fewer with support for bilingual text rendering
2048 x 2048
Max Resolution
Supported Modes
Text to Image
Image Edit
Active
Details
Model ID
z-image-turbo
Tags
image-generation
text-to-image
fast
open-weights
Starting from
$0.0050
/image
via fal.ai
Prices shown are in USD · Some prices estimated from per-megapixel or per-token pricing
Full pricing detailsProvider Performance
Fastest generation through alibaba at 6,041ms median latency with 100.0% success rate.
Aggregated from real API requests over the last 30 days.
Generation Time
alibaba
6,041ms
p95: 10,122ms
Success Rate
alibaba
100.0%
11 / 11 requests
Time to First Byte
alibaba
5,736ms
p95: 12,413ms
Provider Rankings
| # | Provider | p50 Gen Time | p95 Gen Time | Success Rate | TTFB (p50) |
|---|---|---|---|---|---|
| 1 | alibaba | 6,041ms | 10,122ms | 100.0% | 5,736ms |
Data updated every 15 minutes. Based on all API requests through Lumenfall over the last 30 days.
Providers & Pricing (1)
Z-Image Turbo is available exclusively through fal.ai, starting at $0.005/image.
fal.ai
Image Edit
fal/z-image-turbo-edit
Provider Model ID:
fal-ai/z-image/turbo/image-to-image
$0.0050
/megapixel
Z-Image Turbo API OpenAI-compatible
Integrate Z-Image Turbo through Lumenfall’s OpenAI-compatible API to generate images or execute edits using its 6-billion parameter distilled architecture optimized for 8-step inference.
Base URL
https://api.lumenfall.ai/openai/v1
Model
z-image-turbo
Code Examples
Image Edit
/v1/images/editscurl -X POST \ https://api.lumenfall.ai/openai/v1/images/edits \ -H "Authorization: Bearer $LUMENFALL_API_KEY" \ -F "model=z-image-turbo" \ -F "[email protected]" \ -F "prompt=Add a starry night sky to this image" \ -F "size=1024x1024"# Response:# { "created": 1234567890, "data": [{ "url": "https://...", "revised_prompt": "..." }] }import OpenAI from 'openai';import fs from 'fs';const client = new OpenAI({ apiKey: 'YOUR_API_KEY', baseURL: 'https://api.lumenfall.ai/openai/v1'});const response = await client.images.edit({ model: 'z-image-turbo', image: fs.createReadStream('source.png'), prompt: 'Add a starry night sky to this image', size: '1024x1024'});// { created: 1234567890, data: [{ url: "https://...", revised_prompt: "..." }] }console.log(response.data[0].url);from openai import OpenAIclient = OpenAI( api_key="YOUR_API_KEY", base_url="https://api.lumenfall.ai/openai/v1")response = client.images.edit( model="z-image-turbo", image=open("source.png", "rb"), prompt="Add a starry night sky to this image", size="1024x1024")# { created: 1234567890, data: [{ url: "https://...", revised_prompt: "..." }] }print(response.data[0].url)Parameter Reference
Required
Supported
Not available
Core Parameters
| Parameter | Type | Description | Modes |
|---|---|---|---|
prompt
|
string | Required. Text prompt for image generation |
T2I
Edit
|
seed
|
integer | Random seed for reproducibility |
T2I
Edit
|
Size & Layout
| Parameter | Type | Description | Modes |
|---|---|---|---|
size
|
string |
Image dimensions as WxH pixels (e.g. "1024x1024") or aspect ratio (e.g. "16:9")
WxH determines both shape and scale (aspect_ratio and resolution are ignored when size is provided). W:H format is equivalent to aspect_ratio.
|
T2I
Edit
|
aspect_ratio
|
string |
Aspect ratio of the output image (e.g. "16:9", "1:1")
Controls shape independently of scale. Use with resolution to control both. If size is also provided, size takes precedence. Any ratio is accepted and mapped to the nearest supported value.
|
T2I
Edit
|
resolution
|
string |
Output resolution tier (e.g. "1K", "4K")
Controls scale independently of shape. Higher tiers produce larger images and cost more. If size is also provided, size takes precedence for scale. Any tier is accepted and mapped to the nearest supported value.
|
T2I
Edit
|
size
Exact pixel dimensions
"1920x1080"
aspect_ratio
Shape only, default scale
"16:9"
resolution
Scale tier, preserves shape
"1K"
Priority when combined
size
›
aspect_ratio + resolution
›
aspect_ratio
›
resolution
size is most specific and always wins. aspect_ratio and resolution control shape and scale independently.
How matching works
Shape matching
– we pick the closest supported ratio.
Ask for
7:1 on a model with
4:1 and 8:1,
you get 8:1.
Scale matching
– providers use different tier formats:
K tiers (
0.5K 1K 2K 4K)
or megapixel tiers (0.25 1).
If the exact tier isn't available, you get the nearest one.
Dimension clamping
– if a model has pixel limits, we clamp dimensions to fit and keep the aspect ratio intact.
Media Inputs
| Parameter | Type | Description | Modes |
|---|---|---|---|
image
|
file |
Required.
Input image(s) to edit
Supports PNG, JPEG, WebP.
|
T2I
Edit
|
Output & Format
| Parameter | Type | Description | Modes |
|---|---|---|---|
response_format
|
string |
How to return the image
url
b64_json
Default:
"url" |
T2I
Edit
|
output_format
|
string |
Output image format
png
jpeg
gif
webp
avif
Gateway converts to requested format if provider doesn't support it natively.
|
T2I
Edit
|
output_compression
|
integer | Compression level for lossy formats (JPEG, WebP, AVIF) |
T2I
Edit
|
n
|
integer |
Number of images to generate
Default:
1Gateway generates multiple images in parallel even if provider only supports 1.
|
T2I
Edit
|
Additional Parameters
Provider-specific passthrough fields, available only when the request is routed to the listed provider.
| Parameter | Type | Description | Modes |
|---|---|---|---|
|
Universal
|
|||
cfg_scale
|
number | Classifier-free guidance scale — higher values stick more closely to the prompt |
T2I
Edit
|
prompt_enhancement
|
string |
Whether an LLM rewrites/expands the prompt before generation (off, on)
off
on
|
T2I
Edit
|
strength
|
number | How much to transform the input image: 0 keeps it unchanged, 1 fully regenerates from the prompt |
T2I
Edit
|
num_inference_steps
|
integer | Number of inference steps. |
T2I
Edit
|
|
fal
|
|||
acceleration
|
string |
The acceleration level to use.
high
none
regular
|
T2I
Edit
|
enable_safety_checker
|
boolean | If set to true, the safety checker will be enabled. |
T2I
Edit
|
sync_mode
|
boolean | If `True`, the media will be returned as a data URI and the output data won't be available in the request history. |
T2I
Edit
|
Parameter Normalization
How we handle parameters across different providers
Not every provider speaks the same language. When you send a parameter, we handle it in one of four ways depending on what the model supports:
| Behavior | What happens | Example |
|---|---|---|
passthrough |
Sent as-is to the provider | style, quality |
renamed |
Same value, mapped to the field name the provider expects | prompt |
converted |
Transformed to the provider's native format | size |
emulated |
Works even if the provider has no concept of it | n, response_format |
Parameters we don't recognize pass straight through to the upstream API, so provider-specific options still work.
Z-Image Turbo Benchmarks
Alibaba's Z-Image Turbo holds a top 10 position in global text-to-image rankings with a 1253 Elo rating. It remains competitive in high-speed workflows, maintaining a rank of 16 for complex image editing tasks with an Elo of 1021.
Lumenfall Arena
#14
Text-to-Image
1250
Elo
Lumenfall Arena
#20
Image Editing
1110
Elo
Image Editing Landscape
Elo vs Cost
1 model without pricing omitted
Elo vs Speed
6 models waiting for enough speed data
Competition Results
Image Editing
Edit instruction
“Change the scene to night: a deep, dark sky with subtle, glistening stars visible behind the mountain.”
3 attempts – showing best result
Image Editing
Edit instruction
“Give the person a full, thick head of natural hair with realistic texture, density, and a natural hairline. Preserve facial features and lighting.”
Top Matchups
See how Z-Image Turbo performs head-to-head against other AI models, ranked by community votes in blind comparisons.
vs Grok Imagine Image
Challenge: Modern Clean Menu
38% W · 50% L · 13% T
vs Nano Banana
Challenge: Bald man challenge
0% W · 75% L · 25% T
vs GPT Image 1.5
Challenge: Vintage Cafe Logo
0% W · 100% L
vs Seedream 4.5
Challenge: Isometric Miniature Diorama Scenes
0% W · 100% L
vs GPT Image 2
Challenge: Magic Burger Explosion: Fiery Photorealism Challenge
50% W · 50% L
Use Cases
See all Use CasesZ-Image Turbo excels in commercial applications, ranking 5th in product and branding with a 55.6% win rate and 7th in portrait generation with a 62.5% win rate. It demonstrates moderate performance in bilingual text rendering but faces significant challenges in specialized photorealism and portrait-specific arena categories.
Gallery
View all 2 imagesZ-Image Turbo FAQ
How much does Z-Image Turbo cost?
Z-Image Turbo starts at $0.005 per image through Lumenfall. Pricing varies by provider. Lumenfall does not add any markup to provider pricing.
How do I use Z-Image Turbo via API?
You can use Z-Image Turbo through Lumenfall's OpenAI-compatible API. Send requests to the unified endpoint with model ID "z-image-turbo". Code examples are available in Python, JavaScript, and cURL.
Which providers offer Z-Image Turbo?
Z-Image Turbo is available through Replicate, Alibaba Cloud, and fal.ai on Lumenfall. Lumenfall automatically routes requests to the best available provider.
What is the maximum resolution for Z-Image Turbo?
Z-Image Turbo supports images up to 2048x2048 resolution.
Overview
Z-Image Turbo is a 6-billion parameter text-to-image model developed by Alibaba’s Tongyi-MAI team. It distinguishes itself by utilizing distillation techniques to enable high-quality image synthesis in eight steps or fewer, making it significantly faster than standard diffusion models. The model is specifically optimized for bilingual text rendering, supporting both Chinese and English characters within generated imagery.
Strengths
- Inference Latency: By reducing the required sampling steps to a range of 1 to 8, the model provides near-instantaneous image generation suitable for real-time applications.
- Bilingual Text Rendering: The model excels at accurately rendering complex Chinese characters and English text, a task where many Western-centric models often fail or produce “gibberish.”
- Visual Fidelity at Low Step Counts: Despite the aggressive distillation for speed, the model maintains high structural integrity and aesthetic consistency that typically requires 25-50 steps in non-distilled models.
- Multimodal Input Support: It can process both text prompts and image-based references (image-to-image) to guide the generation process, offering flexibility beyond simple text descriptors.
Limitations
- Fine Detail Saturation: While excellent for rapid generation, the model may lack the extreme micro-detail or complex texture depth found in larger, 12B+ parameter models that utilize longer sampling chains.
- Step Count Sensitivity: Moving beyond the 8-step threshold does not necessarily improve quality and can sometimes lead to visual artifacts, as the model is strictly tuned for low-step schedules.
- Stylistic Range: Compared to broader foundation models, the output may lean toward a specific “polished” aesthetic favored by its distillation process, which might require more aggressive prompting to deviate from.
Technical Background
Z-Image Turbo is part of the Z-Image model family and utilizes a distilled architecture derived from a larger latent diffusion framework. To achieve its speed, the developers employed a consistency-based distillation approach that maps the probability flow of the original model into a single or few-step inference trajectory. The integration of a specialized text encoder allows the model to handle bilingual tokens more effectively than models trained solely on English datasets.
Best For
This model is ideal for interactive applications such as live drawing assistants, rapid prototyping for UI/UX design, and social media content creation where speed is prioritized over granular control. It is also a leading choice for projects requiring accurate Chinese typography within images. Z-Image Turbo is available for integration and testing through Lumenfall’s unified API and interactive playground.
Try Z-Image Turbo in Playground
Generate images with custom prompts — no API key needed.