# FLUX AI by Black Forest Labs: FLUX.2 and FLUX 3 Video

> FLUX AI is a family of models from the German lab Black Forest Labs. FLUX.2 (November 25, 2025) generates and edits images up to 4 MP, takes up to 10 references and renders text and infographics, with open weights for the dev version. FLUX 3 Video (August 4, 2026) makes clips up to 20 seconds with sound and speech and ranks #3 on Arena's video leaderboard. FLUX is available in the AI2Face generator.

Developer: Black Forest Labs · Released: 2025-11-25 · Updated 2026-10-02
URL: https://ai2face.com/models/flux

## Key facts

| | |
|---|---|
| Developer | Black Forest Labs (Germany) |
| Versions | FLUX.2 — pro, flex, max, dev, klein (November 25, 2025); FLUX 3 Video (August 4, 2026) |
| What it does | FLUX.2 — image generation and editing; FLUX 3 Video — video with sound |
| Resolution | FLUX.2 — up to 4 MP; FLUX 3 Video — HD, Full HD via upscaling |
| References | FLUX.2 — up to 10 images at once |
| Video | Up to 20 seconds with sound, dialogue and lip sync in many languages |
| Open weights | FLUX.2 [dev] with 32 billion parameters, FLUX.2 [klein] — Apache 2.0 |
| In AI2Face | Yes |

## What is FLUX AI

FLUX is a family of models from the German lab Black Forest Labs (BFL). The original FLUX.1 became one of the best-known open image models, and many FLUX reviews still describe it. In 2026, two other models are the current ones.

**FLUX.2** came out on November 25, 2025, and brought image generation and editing together. Variants:

- **[pro]** — the default version, with quality on par with the best closed models;
- **[flex]** — adjustable steps and guidance, strong on text and fine detail;
- **[max]** — the top-tier version;
- **[dev]** — open weights with 32 billion parameters on Hugging Face;
- **[klein]** — a compact model under the Apache 2.0 license.

BFL describes **FLUX 3** as a single multimodal model for video with sound, images and action prediction. FLUX 3 Video came first, on August 4, 2026. FLUX 3 Image and an open FLUX 3 Dev are planned as the next stages.

## What FLUX can do

**Photorealism.** FLUX.2 [pro] handles skin, light and textures with confidence — the potter portrait in the gallery is a first try.

**Exact colors.** You can specify a color as a hex code in the prompt, and FLUX.2 will match it — the vase background in the gallery is set to #E8DCC8. For brand assets and product listings, that saves retouching.

**Text and layout.** BFL claims reliable rendering of complex typography, infographics, memes and UI mockups with small text. The COAST OF LIGHT headline on the gallery poster came out without errors.

**Up to 10 references and 4 MP.** FLUX.2 takes up to 10 images into account at once — faces, products, style — and edits images at up to 4 megapixels.

![On the left, two references — a gray-green vase with pampas grass and a ceramicist in an apron at a pottery wheel; on the right, the same ceramicist holding that vase in her workshop](https://ai2face.com/_astro/multi-reference.DBw3yizi_2bHv72.jpg "Two references in one prompt: the face and apron from one image, the vase from another. FLUX.2 [pro], 2K")

**Video with sound.** FLUX 3 Video makes clips up to 20 seconds in HD, with Full HD via upscaling. Sound, dialogue and effects are generated together with the picture, and lip sync works in many languages, Russian included. There are text-to-video, image-to-video and keyframe modes, clip extension and multiple shots in one video. Draft mode returns a quick preview, and the final render uses the same prompt.

[A hot-air balloon festival at dawn over a misty valley: dozens of colorful balloons rise as the camera flies between them, with a burner roaring in the foreground](https://ai2face.com/_astro/media/models/flux/balloons.mp4 "The camera flies between balloons to the sound of a burner. FLUX 3 Video, 720p, 6 s")

## What FLUX is good for

- **Product listings and brand graphics** — exact hex colors and clean product lighting.
- **Posters, infographics, mockups** — text and fine typography.
- **Series with one character or product** — up to 10 references.
- **Your own infrastructure** — the open FLUX.2 [dev] and [klein] can be fine-tuned and self-hosted.
- **Short clips with speech** — FLUX 3 Video with multilingual lip sync.

## FLUX vs. alternatives

Images:

| | FLUX.2 | [Nano Banana 2](/models/nano-banana-2) | [Seedream](/models/seedream) 5.0 Pro | [GPT Image](/models/gpt-image) 2.5 |
|---|---|---|---|---|
| Arena rank (Sep 24, 2026) | 27–30 (max, flex, pro) | 9 | 10 | 1 |
| Resolution | up to 4 MP | up to 4K | up to 2K | up to 3840 px |
| References | up to 10 | up to 14 | up to 10 | several |
| Hex colors | yes | — | yes, edits by color code | — |
| Open weights | yes ([dev], [klein]) | no | no | no |

Video:

| | FLUX 3 Video | [Gemini Omni](/models/gemini-omni) | [Seedance](/models/seedance) 2.5 | [Wan 3.0](/models/wan) |
|---|---|---|---|---|
| Arena rank (Sep 21, 2026) | 3 | 1–2 | 7 | 6 |
| Duration | up to 20 s | 3–10 s | up to 30 s | up to 30 s |
| Sound and speech | yes | yes | yes | yes |

For images, FLUX.2 trails the leaderboard leaders but wins on open weights and exact colors. For video, FLUX 3 went straight into the top three.

## How to write prompts for FLUX

- **Describe the whole scene instead of listing tags.** FLUX.2 understands coherent sentences better than “portrait, 8k, masterpiece.”
- **Give colors in hex** when the brand matters: “background in exact color #E8DCC8.”
- **Put text in quotes** and add “no other text.”
- **Assign roles to references:** “the face from the first image, the jacket from the second.”
- **For video,** start in draft mode, then run the final render with the same prompt.

## How to use FLUX in AI2Face

FLUX is available in the AI2Face generator, no BFL account needed. You can compare FLUX.2 with [Nano Banana](/models/nano-banana) and Seedream, or FLUX 3 Video with Gemini Omni and Seedance, on the same prompt. The cost of each generation is shown before you run it, you can pay by card or crypto, and credits from packs don't expire. New accounts get welcome credits after signing up.

## Limitations

- **FLUX 3 Image isn't out yet.** For images, BFL's current model is still FLUX.2.
- **FLUX.2 ranks below the image leaders** — 27th–30th place versus 1st–10th for GPT Image, Nano Banana and Seedream.
- **Full HD video comes from upscaling.** FLUX 3 Video's base resolution is HD.
- **Video editing and Omni Reference** in FLUX 3 are promised by BFL for later.
- **The open models need serious hardware:** FLUX.2 [dev] has 32 billion parameters.

## Bottom line

FLUX AI in 2026 is no longer the FLUX.1 of older reviews. FLUX.2 is a strong model for product listings, posters and brand graphics, with exact colors and open weights, while FLUX 3 Video broke into the top three video models with clips up to 20 seconds and multilingual speech. For maximum image quality, use GPT Image or Nano Banana; for brand colors and your own infrastructure, use FLUX.2; for video, compare FLUX 3 with Gemini Omni in AI2Face.

## Generation examples

- A woman potter in a sunlit workshop with clay on her hands, shelves of ceramics behind her
  Prompt: `Photorealistic portrait of a 45-year-old woman potter in her sunlit workshop, clay on her hands and apron, shelves of ceramic bowls behind, soft window light, natural skin texture, 85mm lens, shallow depth of field, no text`
- A matte sage-green vase with dried pampas grass on a beige plinth
  Prompt: `Minimal product shot of a matte sage-green ceramic vase with dried pampas grass on a beige plinth, background in exact color #E8DCC8, soft shadow, studio light, premium catalog style, no text, no logos`
- An art deco COAST OF LIGHT poster: a white lighthouse on rocks and sailboats
  Prompt: `Vintage travel poster in art deco style: a white lighthouse on a rocky coast with sailboats, bold flat shapes, limited palette of navy, coral and cream, headline text COAST OF LIGHT at the top in elegant art deco letters, no other text`
- An isometric illustration of a tiny greenhouse café with plants, a barista and customers
  Prompt: `Cozy isometric illustration of a tiny greenhouse cafe with plants, a barista, two customers at small tables, warm light, soft pastel colors, clean 3D render style, high detail, no text`

## FAQ

### What is FLUX AI?

A family of models from the German lab Black Forest Labs. FLUX.2 generates and edits images, and FLUX 3 Video makes clips with sound. The original FLUX.1 became one of the most popular open image models.

### How is FLUX.2 different from FLUX.1?

FLUX.2 combines generation and editing in one model, takes up to 10 references at once, works at up to 4 MP and reliably renders complex typography, infographics and UI mockups with small text.

### Can I download FLUX?

Partly, yes. FLUX.2 [dev] with 32 billion parameters is on Hugging Face, and the compact FLUX.2 [klein] is released under the Apache 2.0 license. The pro, flex and max versions and FLUX 3 Video run via API.

### Can I use FLUX for free?

The open FLUX.2 [dev] and [klein] are free if you run them on your own machine, but you'll need a powerful GPU. In AI2Face, you can try FLUX with the welcome credits you get after signing up.

### Can FLUX make videos?

Yes. Since August 4, 2026, FLUX 3 Video makes HD clips up to 20 seconds with sound and dialogue in many languages. It ranks #3 out of 48 video models on Arena.

## Sources

- https://bfl.ai/blog/flux-2
- https://bfl.ai/blog/flux-3
- https://bfl.ai/blog/flux-3-video
- https://bfl.ai/pricing
- https://help.bfl.ai/articles/7655484417-what-flux-models-are-available
- https://arena.ai/leaderboard/text-to-image
- https://arena.ai/leaderboard/text-to-video