Create On-Brand Visuals With Flux 2: Exact Colors, Readable Text
Last verified: November 3, 2026
Flux 2 is the image generation and editing family from Black Forest Labs (BFL) that follows FLUX.1. Every model in the family supports precise hex color control and reliable text rendering , so you can write a brand palette and a headline straight into the prompt for posters, packaging and ads. BFL released the family on November 25, 2025 and added the FLUX.2 [klein] models, which generate in under a second, on January 15, 2026 . FLUX.2 uses a single latent flow matching model for both generating and editing. Inside it, the Mistral-3 24B vision-language model supplies real-world knowledge and context, and a rectified flow transformer handles spatial relationships, materials and composition .
FLUX.1 showed what media models could do as creative tools. FLUX.2 is built for production work. It keeps characters, products and styles consistent across multiple reference images, follows structured prompts, sticks to brand guidelines and handles lighting, layouts and logos reliably. BFL retrained the latent space from scratch for better learnability and higher image quality, and says this generation has substantially better prompt adherence, world knowledge and typography. Within the family, FLUX.2 [max] produces the highest-quality output, with the strongest prompt following and the most consistent edits across tasks . On this page you can pick Flux 2 Pro, Flux 2 Flex or Flux 2 Max, each in text to image and image to image.
Explore Flux AI's Models
Specs at a Glance: Pro, Flex and Max
What each version gives you on this page: output size, frame shape, prompt room and reference uploads.
Images per run
One image each time you click Generate
Output resolution
1K or 2K on Pro and Flex · 1K, 2K or 4K on Max · 1K default
Frame shapes
Seven ratios from 9:16 to 16:9 on every version · Max adds 21:9, 9:21 and Auto (editing only)
Prompt capacity
2,048 characters on Pro and Flex · 5,000 on Max
Reference images per edit
1 to 8 photos combined in one Image to image run
Accepted reference files
JPG, JPEG, PNG or WEBP · up to 10 MB · at least 300×300 px
Before You Hit Generate
Six checks that help colors, copy and references come out the way you planned.
Choose the frame before you compose
Decide where the image will be used first. Pro and Flex cover square, portrait and landscape up to 16:9. For 21:9 banners or 9:21 story frames, switch to Flux 2 Max, then describe the composition for that shape, for example a subject on the left third of a wide frame.
Attach each hex code to a named element
FLUX.2 reads color values written as hex in the prompt, such as 'color #FF5733' . Tie each code to one surface, like 'label band in #0B3D91', instead of listing a loose palette, so every color has a clear target.
Spell out on-image copy in quotes
For headlines, labels and signage, write the exact words in quotation marks and give each line its own position and type style, for example a bold condensed sans-serif headline across the top. Short, separate lines are easier to place than one long block of copy.
Budget characters for structured briefs
Flux 2 Pro and Flux 2 Max accept JSON-style prompts for complex scenes , but braces and keys count toward the limit: 2,048 characters on Pro, 5,000 on Max. Keep only the fields that matter, such as scene, subject, text, lighting and camera, or move long briefs to Max.
Prepare and number your references
Each upload must be a JPG, JPEG, PNG or WEBP file of at least 300×300 px and no more than 10 MB, with eight files at most. In the prompt, refer to uploads by their order. FLUX.2 handles phrasing like 'the woman in image 1 standing in the pose of image 2' .
Name the camera for photoreal shots
For photographic realism on Flux 2 Pro and Flux 2 Max, add a camera reference: the body, lens and settings, such as 'shot on a medium-format camera, 80mm, f/2.8' . Add the light source and its direction so the scene reads as one real setup.
Flux 2 or Nano Banana 2: Pick the Right Generator for the Job
Flux 2 Pro, Flex and Max and Nano Banana 2 each run in text to image and image to image. The first six rows show the settings each generator gives you here. The last two describe the models themselves, as Black Forest Labs and Google present them.
| Feature/Spec | Flux 2 | Nano Banana 2 |
|---|---|---|
| Modes you can run | Text to image and Image to image, on Flux 2 Pro, Flux 2 Flex and Flux 2 Max | Text to image and Image to image |
| Resolution choices | Flux 2 Pro and Flux 2 Flex: 1K or 2K · Flux 2 Max: 1K, 2K or 4K (1K default on all three) | 1K, 2K or 4K (1K default) |
| Aspect ratio choices | Flux 2 Pro and Flux 2 Flex: 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9 · Flux 2 Max: the same seven plus 21:9 and 9:21, with Auto in Image to image | 1:1, 2:3, 3:2, 3:4, 4:3, 9:16, 16:9 |
| Prompt length per run | Flux 2 Pro and Flux 2 Flex: 2,048 characters · Flux 2 Max: 5,000 characters | 2,048 characters |
| Reference images (Image to image) | 1–8 images on Flux 2 Pro, Flux 2 Flex and Flux 2 Max; JPG, JPEG, PNG or WEBP; up to 10 MB each; at least 300×300 px | 1–8 images; JPG, JPEG, PNG or WEBP; up to 10 MB each; at least 300×300 px |
| Google Search setting (Nano Banana 2, Text to image) | The Flux 2 Pro, Flux 2 Flex and Flux 2 Max generators show no such setting | Google Search: No or Yes, default No |
| Creator and release | Black Forest Labs; FLUX.2 family released November 25, 2025 ; FLUX.2 [max] added December 16, 2025 | Google DeepMind; released February 26, 2026 as Gemini 3.1 Flash Image |
| Model foundation | Latent flow matching architecture coupling the Mistral-3 24B vision-language model with a rectified flow transformer; one model generates and edits | Gemini 3.1 Flash Image, bringing high-fidelity generation and faster advanced editing to Google's Flash model |
Where the Two Generators Pull Apart
Canvas size and frame shape
At standard ratios, the choice comes down to pixel count. Flux 2 Pro and Flux 2 Flex suit social posts, web graphics and drafts that don't need more than 2K. Large prints, layouts you plan to crop heavily and hero images point to Flux 2 Max or Nano Banana 2, which both reach 4K. Wide cinema-style banners, website headers and tall phone-story frames are different. Of the two generators, only Flux 2 Max offers those ultrawide and ultra-tall shapes, so to compose for them directly instead of cropping a 16:9 or 9:16 result, you need Max.
Brief length and live web context
Flux 2 Max's prompt box holds more than twice as much text as the others. That leaves room for long, itemized briefs that list every subject, where it sits, the lighting and each line of copy. On Flux 2 Pro and Flex the limit matches Nano Banana 2's, so prompt length only sets Max apart. Nano Banana 2 serves a different need: in text to image it has a Google Search toggle, so a generation can draw on current web information. That helps with visuals tied to recent events, live products or real places. The Flux 2 generators on this page have no comparable setting. For edits from reference photos, both sides accept exactly the same uploads, so that step won't decide it.
Which Generator Fits Your Next Image?
Use this quick guidance to pick the best option for your workflow.
How to choose: Pick Flux 2 Max when the frame has to be ultrawide or ultra-tall. It adds 21:9, 9:21 and, for edits, Auto to the seven ratios Nano Banana 2 offers. Max is also the choice when your brief runs past 2,048 characters, since it accepts up to 5,000. Pick Nano Banana 2 when a text-to-image result should draw on Google Search, a setting the Flux 2 generators don't show, or when you need 4K and would otherwise use Flux 2 Pro or Flex, which stop at 2K. Several things are the same on both sides: Flux 2 Max and Nano Banana 2 both reach 4K, every version runs text to image and image to image, and both accept the same one to eight reference files, so none of these will decide it for you.
From Prompt to Finished Image in Five Steps
Five quick steps, from picking a version to generating your image.
Choose a version in the model list
Select Flux 2 Pro, Flux 2 Flex or Flux 2 Max, then a mode. Use Text to image to start from words, or Image to image to work from your own photos.
Upload photos with Add Image
In Image to image, add between one and eight JPG, JPEG, PNG or WEBP photos. Each must be at least 300×300 pixels and no larger than 10 MB. Text to image skips this step.
Write in the Prompt box, or use Enhance Prompt
Describe the scene, colors and any words that should appear in the image. You get up to 2,048 characters on Pro and Flex and 5,000 on Max. The Enhance Prompt button under the box suggests fuller versions of a rough idea, and your text only changes if you pick one.
Set Aspect Ratio and Resolution
Pick the frame shape, then the output size. Pro and Flex have a Resolution setting with 1K or 2K. Max has Resolution Quality with 1K, 2K or 4K. All three start at 1K.
Click Generate
Start the run to get your image. Each click on Generate produces one image.