Black Forest Labs

Flux — The Model That Leads on Photorealism and Text

The strongest open-weight family for photorealistic people and for text rendered inside images.

Short answer

Flux came out of Black Forest Labs, founded by some of the original Stable Diffusion developers, and quickly became the reference in two areas: photorealistic images of people, and legible text rendered inside an image.

Those two areas were long the weakest points of AI image generation. Broken hands and meaningless piles of letters were the two details that gave away an image as "AI work". Flux largely solved both.

The confusing part is distribution: Flux has no single official site or interface. Three different versions are offered under different licences through different platforms. Before using it commercially you must know which version you are running.

Strengths

  • A clear lead on photorealistic people — hands and faces come out consistent
  • One of the few models that renders legible text inside an image
  • Follows long, detailed prompts faithfully
  • Open-weight versions can run on your own hardware or in the cloud

Weaknesses

  • There is no single official interface; you use it through various platforms, which is confusing
  • Licences differ by version — the dev version is not licensed for commercial use
  • Less striking than Midjourney on artistic, stylised work
  • Running it locally demands more VRAM than Stable Diffusion

Three versions, three different licences

This is the one thing to understand before using Flux, and it is where many users go wrong.

Three versions, three different licences
VersionCommercial useHow to accessWhen to choose it
SchnellAllowed (open licence)Free platforms, local installFast drafts and free generation
DevNot allowed — personal onlyLocal install, cloud servicesPersonal projects and experiments
ProAllowed (paid API)API and partner platformsCommercial work, highest quality

Rendering text inside an image

This is Flux's most practical advantage. When a poster, package or social image needs legible text, most models produce letter-like scribbles while Flux actually writes.

Method matters: keep the text in quotes and short. A clear instruction like "a sign that reads 'NOW OPEN'" is far more reliable than a long sentence. Beyond three words the error rate climbs quickly.

With Turkish characters the picture is mixed: ç, ö and ü usually come out right, while ğ, ş and ı sometimes break. For critical text such as a brand name, the safest route is to leave that area empty and add the lettering afterwards in a design tool.

Pricing

Flux comes in three versions with different licences: Schnell is openly licensed and commercially usable, Dev is non-commercial only, and Pro is paid through the API. Before using it commercially, verify which version you are actually running.

Information verified on:

Frequently asked questions

Can Flux be used for free?

The Schnell version is openly licensed and can be tried free on many platforms; installed locally it is free indefinitely. The Dev version downloads free but is not licensed for commercial use. The highest-quality Pro version is paid through the API.

Is Flux or Midjourney better?

It depends on the job. For photorealistic people, product photography and designs containing text, Flux is better. For artistic, stylised and atmospheric imagery, Midjourney is more striking. Choose Flux when it must look like a photograph and Midjourney when it must look like a painting.

Which Flux version can I use commercially?

The Schnell and Pro versions are cleared for commercial use. The Dev version is not — it is licensed for non-commercial use only. If you are using it through a platform, verify which version it runs; many interfaces do not state this clearly.

Prompts for this tool

Corporate headshot

professional headshot of a woman in her forties, plain charcoal blazer, soft diffused window light from the left, neutral grey seamless background, shot on 85mm f/2, shallow depth of field, natural skin texture, calm confident expression, catchlight in the eyes

Works well in these tools: flux, stable-diffusion, midjourney

Note: The phrase "natural skin texture" is critical: without it models over-smooth skin and the result looks plastic.

Candid portrait in natural light

candid portrait of an older man laughing, seated by a cafe window, warm late afternoon sunlight raking across his face, blurred street behind, shot on 50mm f/1.8, film grain, muted warm colour grade, documentary photography

Works well in these tools: flux, midjourney

Note: The words "candid" and "documentary photography" break the posed stock-image feel.