FLUX.2

Overview

FLUX.2 Text-to-Image is the text-only generation pipeline of the image generation model released by Black Forest Labs, called through the unified HiAPI.ai interface. It takes a text prompt and outputs photorealistic images, and exposes parameter controls such as resolution, aspect ratio, output format and quality, and random seed, making it suitable for creative concept exploration and batch visual production scenarios that require stable, reproducible image output.

Key Highlights

• Configurable resolution and aspect ratio — Supports output resolutions from 0.5 to 4 MP and a variety of common aspect ratios, as well as custom width and height, so you can choose sharpness and composition according to use.
• Reproducible generation via random seed — Setting a seed reproduces the same result, making it easy to compare parameters and run version regression on the same composition.
• Controllable output delivery parameters — Choose webp, jpg, or png format and adjust compression quality to directly produce distributable image files.
• Adjustable safety tolerance — safety_tolerance offers a range of 1 to 5, letting you adjust filtering strictness to meet business compliance requirements.

Capabilities

Text-to-photorealistic image generation: Enter a text prompt to generate an image; official and third-party materials both position it primarily for photorealistic style, usable for concept visuals and asset production.

Aspect ratio and resolution control: Select a preset ratio via aspect_ratio or customize width and height, paired with resolution parameters from 0.5 to 4 MP to match landscape or portrait compositions and different display placements.

Output format and quality adjustment: Supports three output formats — webp, jpg, and png — and allows setting save quality from 0 to 100 to balance file size and image quality according to the use of the output.

Reproducible generation: After specifying a seed, the same prompt and parameter combination yields consistent results, making it easy to identify the source of changes during iteration.

Safety filtering adjustment: The safety_tolerance parameter provides five levels from strictest to most lenient, which can be set according to the compliance standard of the launch scenario.

Use Cases

• Concept sketches for creative teams — Quickly produce multiple visual drafts with short prompts for direction screening and internal review, then decide what to refine.
• Composition exploration for design and advertising — Generate landscape or portrait visuals using aspect ratio and resolution parameters, directly usable for proposal drafts and layout previews.
• Batch asset production pipelines — Use fixed seeds and output format parameters to batch-generate image assets with uniform specifications in automated workflows.
• Photorealistic visual references — Generate photorealistic scenes or product mood images as reference material before photography, set design, or rendering.

Limitations and Selection Advice

This pipeline accepts only text prompts and does not process reference images; when you need to edit an existing image or perform style transfer, use the image-to-image pipeline in the same family instead. Resolution support goes up to 4 MP, but the parameter documentation recommends 2 MP and below; for very large output sizes, evaluate the image quality gain yourself. Safety filtering applies by level, and stricter levels may affect the visibility of generated content. The non-commercial license and hardware deployment requirements mentioned in research materials depend on how users actually obtain the model; when calling through HiAPI.ai, such self-deployment constraints can be ignored.