Imagefluxflux-dev

FLUX.1 [dev]

12B-parameter flow transformer for high-quality text-to-image. Strong prompt adherence, competitive with closed-source alternatives.

Open in Image Lab Browse all models~$0.05 per run

Flux guideVerified

Black Forest Labs' open-weights image model line. Flux 2 Pro is the current flagship — best-in-class prompt adherence and photorealism. Flux Kontext (separate brand below) handles editing.

Strengths

Best prompt adherence in open-weights image models.
Photorealism: skin, fabric, lighting, depth all look correct.
Handles long, dense prompts without losing the lead subject.
Flux 2 Pro is the picker default for finals.

Weaknesses

Text rendering inside images is OK but not as good as Ideogram.
Stylization (painterly, anime) needs explicit prompting — defaults to photoreal.

Best for

Photoreal portraits, product shots, environments
Hero stills for video pipelines (Flux still → Runway video)
Anything that needs lens vocabulary (35mm, anamorphic, shallow DoF)

Prompting tips

Lead with lens / camera specs ("35mm anamorphic, shallow depth of field") — Flux loves this language.
Lighting terms work: "rim light", "soft key", "chiaroscuro".
Be specific about the subject's positioning — Flux respects spatial prompts.

Parameters

image_size
string
No description.
square_hdsquareportrait_4_3portrait_16_9landscape_4_3landscape_16_9
default: landscape_4_3
num_inference_steps
int
More steps = higher quality, slower generation.
default: 28
range: 1 … 50
guidance_scale
number
How strictly to follow the prompt (3.5 = balanced).
default: 3.5
range: 1 … 20
num_images
int
No description.
default: 1
range: 1 … 4
seed
int
Set for reproducible output. Leave empty for random.
enable_safety_checker
boolean
No description.
default: true
output_format
string
No description.
jpegpng
default: jpeg

You'll need

A text prompt

Try now

More from flux