Home/Large models/Image generation

Image generation

Text-to-image, image-to-image and inpainting.

Style locking, subject consistency, upscaling and compliance checks run on one path, so the images that come out are ready for your asset library.

1.8s
P50 for a single 1024px image
4096px
Maximum output resolution
60+
Built-in reproducible style presets
99.7%
First-shot usable rate
Overview

Reproducible image generation at batch scale

Anchor a character or product once and later compositions keep it consistent. A tuned parameter set can be saved as a named style for the whole team. Every generation passes a quality and compliance check, and anything that fails is re-run without using quota.

1.8s
P50 for a 1024px image
99.7%
First-image usable rate
4 min
Time to produce 1,000 product images
Advantages

Built for production

01

Subject consistency

Anchor a character or product once and keep the same face and packaging across later compositions.

02

Reproducible styles

Use the built-in presets, or save your own tuned parameter set as a named style.

03

High-resolution output

Built-in upscaling and detail repainting keep edges clean and text intact.

04

Automatic checks

Quality, copyright risk and content compliance are checked automatically, and failures are re-run without being billed.

Capabilities

API capabilities

Text-to-image, image-to-image, inpainting

Three modes on one endpoint, with masks specified at pixel or semantic level.

Size and composition control

Any aspect ratio, with pose, line-art and depth-map structural constraints.

Batch jobs

Submit several prompts at once, scheduled concurrently and returned by asynchronous callback.

Negative prompts and fixed seeds

The same seed reproduces an earlier result exactly, which helps A/B tests and debugging.

Clear commercial rights

The commercial scope of each result comes back with the response, with no legal grey area.

Straight to object storage

Results can be written to your bucket and CDN without a round trip.

Product and packaging assets

Shoot a product once and derive the full set of scene shots, colour variants and store-listing sizes.

In-app scene illustrations

Illustrations for rooms, scenes and device cards produced as a set in one style.

Manuals and installation diagrams

Installation steps, button legends and pairing guides drawn to one visual standard.

Localised market assets

Swap setting and people per region for the same selling point while the product itself stays identical.

img = starsg.image(
    model="bfl/flux-pro-1.1",
    prompt="纯白背景上的哑光陶瓷杯,柔光,居中",
    style="ecommerce-clean",   # 命名风格,可复现
    subject_ref="sub_c81f...",  # 锁定同一款商品
    size="2048x2048",
)
print(img.url, img.usage.cost_usd)
Integrate

Anchor IDs keep it consistent

Register a product or character as an anchor ID and point every later generation at it — series consistency comes for free.

Python / Node / Go SDKOne API key
Read the integration docs

Start building with Image generation

Sign up to create an API key. It shares the same auth, billing and observability as the other six capabilities.