You have a product shot that is almost right. The bottle sits in the wrong spot, the cup on the table should be something else, and the sign on the wall needs your brand name. Normally that means a brand new generation and a lot of hoping that the good parts survive. Flux 3 is now live on Atlabs, and it is built around fixing exactly this kind of problem. You can place objects where you want them, edit an image again and again without disturbing the rest of it, and blend up to 10 references into one result. This guide shows the real prompts and the real outputs, so you can see what each control does before you try it yourself.
Why Flux 3 Matters for Creators and Marketers
Most image models ask you to describe a scene and accept whatever comes back. That works for a first idea, but it breaks down the moment you need a specific layout, a specific product position, or a small change to an image you already like. Creators who build ads, thumbnails, storyboards and product mockups hit this wall every day, because the last ten percent of an image is where the real work sits.
Flux 3 goes after that last ten percent. The launch video shows a photoreal portrait that holds up down to every pixel, which matters when your image will be cropped, upscaled or placed on a large screen. It also shows three controls that change how you work. Layout control lets you decide where each object goes. Multi step edits let you change one detail at a time while the rest of the image stays locked. Reference blending lets you feed in up to 10 images and get a single coherent frame back, including a native 4K result at 5456 × 3072 pixels.
The Atlabs Workflow for This
On Atlabs, Flux 3 lives in the Create Image app, the single step tool where you generate a still before you build anything else. It sits next to the other image models in the same workspace, including Nano Banana, so you can try the same idea in more than one model without opening a second account. You pick the model, write your prompt, and generate.
The still is only the starting point. Once you have a frame you like, you can send it to Image to Video, which is powered by Google Veo 3.1 Fast, and animate it from a start frame. You can also use Reframe Image to resize it for 9:16, 16:9 or 1:1, and Upscale Image to sharpen it before it goes out. The whole path from first prompt to finished asset stays inside one platform.
Step-by-Step Tutorial: Use Flux 3 in Atlabs
This walkthrough follows the same order as the launch video: layout first, then edits, then references, then output. Each step uses a real prompt and a real result.
Open Create Image and choose Flux 3. From the Atlabs app dashboard, open the Create Image app and select Flux 3 as your model. Choose your aspect ratio before you generate, because the shape of the frame affects where you place objects.
Set the layout. Decide what goes where before you generate. In the launch example, three objects are placed on the canvas with labeled boxes: a perfume bottle on the left, a chrome sphere in the middle, and 3 peaches on the right. Each box carries a plain label with a count where it matters, and Flux 3 places each object inside its box.
Add references if you need them. If you want a specific face, product or style, add reference images. Flux 3 accepts up to 10 references in one generation, and the launch video shows a native 4K result made from 5 of them.
Generate, then edit in steps. Review the first result, then edit it with a short instruction. Edit one detail at a time. In the launch example, the first edit swaps a coffee cup for a strawberry milkshake, and the second edit changes the neon sign on the wall. After both edits, 94.7% of the pixels in the image were untouched.
Check the result at full size. Zoom in on the details that matter for your use case, such as text on a sign, the edge of a product, or a face. If something is slightly off, run one more small edit rather than starting over.
Send the still to the next step. Use Image to Video to animate it, Reframe Image to fit another platform, or Upscale Image for extra sharpness.
Watch the Tutorial
This Atlabs tutorial walks through advanced prompting for visuals, which pairs well with the Flux 3 prompts below.
Flux 3 Prompting Guide with Real Outputs
Every prompt and image below comes from the Flux 3 launch video on Atlabs. Nothing here is mocked up.
Prompt 1: Layout control
Instead of describing positions in a long sentence, you draw a box for each object and label it. This is the layout used in the launch video, from left to right.
Output: a pink glass perfume bottle, a mirror finish chrome sphere and three stacked peaches, each placed inside its own box on a pink terrazzo slab against a warm cream wall. The count in the label ("3 peaches") is respected, and the chrome sphere reflects the scene around it.
Try this prompt in Atlabs Create Image.
Prompt 2: First edit
Start with the image you already have. The starting frame shows a woman in a pink coat sitting at the counter of a retro diner, with a coffee cup on the counter. The edit prompt is one short instruction.
Output: the coffee cup is gone and a strawberry milkshake with a cherry on top sits in its place. The woman, the counter, the lighting and the diner behind her stay the same.
Try this prompt in Atlabs Create Image.
Prompt 3: Second edit on the same image
Now edit the result again. The blank sign on the wall is the next target.
Output: the blank sign on the wall now glows pink and reads FLUX 3, and the milkshake from the first edit is still there. Across both edits, 94.7% of the pixels in the image were left untouched, which is why the second change did not undo the first.
Try this prompt in Atlabs Create Image.
Prompt 4: Reference blending
The last example uses references instead of a text edit. Flux 3 can blend up to 10 reference images, and the launch video shows one result made from 5 of them.
Output: a corgi in a blue bellhop uniform and orange sunglasses, standing in a pink hotel lobby with a luggage cart and a balloon behind it. The frame is native 4K at 5456 × 3072 pixels, and the character, the setting and the color palette all come from the references rather than from a long text prompt.
Try reference blending in Atlabs Create Image.
Why Atlabs Works Well for Flux 3
Atlabs puts Flux 3 in the same workspace as the tools you use next. A frame made in Create Image can go straight into Image to Video, where Google Veo 3.1 Fast animates it from a start frame. If a clip needs more motion or a different look, the model choice stays open, since Atlabs carries other models such as Kling 3.0 and Seedance 2.0 for video work.
The frame level controls matter here too. Reframe Image resizes a still to 9:16, 1:1 or 16:9 without a trip to another app, and Upscale Image adds sharpness before export. For a creator who builds ads or social content, that means one asset can become a vertical story, a square post and a wide thumbnail from the same Flux 3 image.
Pro Tips for Better Flux 3 Results
Change one thing per edit. Both real edits above are a single short instruction with one target, a cup in the first and a sign in the second. Small, specific edits are easier to check and easier to undo than one long instruction that asks for four changes.
Put the count in the label. The layout example uses "3 peaches" rather than "peaches", and the output shows three. When quantity matters, say it in the box label.
Build your references like a casting call. If you blend several references, choose images that each add one clear thing, such as a character, a location or a color mood. Ten crowded references give the model less to work with than five clean ones.
FAQ
What is Flux 3 on Atlabs?
Flux 3 is an image model that is now live on Atlabs. It focuses on photoreal output, layout control, multi step edits that leave the rest of the image untouched, and blending up to 10 references into one result.
How many reference images can I use?
Flux 3 can blend up to 10 references in one generation. The launch video shows a native 4K result made from 5 references.
What resolution can Flux 3 produce?
The native 4K example in the launch video is 5456 × 3072 pixels. For extra sharpness on any image, you can also run it through the Upscale Image app in Atlabs.
Can I turn a Flux 3 image into a video?
Yes. Use the Image to Video app, which is powered by Google Veo 3.1 Fast. Upload your image as a start frame, describe the motion, and generate.
Final Verdict
Flux 3 gives image creators control over the three things that usually cost the most time: where objects sit, what changes between versions, and how references shape the result. The real prompts above show it clearly. A layout drawn with boxes landed in the right places, two edits on one diner image left nearly every other pixel alone, and five references produced a native 4K frame.
On Atlabs, that control sits in the same place as Image to Video, Reframe Image and Upscale Image, so a single Flux 3 frame can become an animated clip, a vertical post and a sharpened asset without leaving the platform. If you make ads, thumbnails, storyboards or product visuals, Flux 3 on Atlabs is worth a test with your own layout, your own edits and your own references.
