A useful image workflow continues after the first generation. You might like the room but want a different chair, need a warmer light, or want to replace a product with a specific reference. FLUX 3 Image brings image creation and editing to Kolbo.AI, while Magic Edit gives you a canvas for explaining changes where they belong.
FLUX 3 Image is an image model family, distinct from the FLUX 3 video models. Choose its image creation option when starting from a prompt, or its editing option when working with an existing image.
The demo presents FLUX 3 Image editing results in an animated film: it illustrates still-image edits, rather than a video editing feature. Give each change a place, an instruction and, when useful, its own reference.
Create the Starting Image
Open image creation and select FLUX 3 Image. Describe the scene, main subject, composition and lighting you want. A practical prompt might ask for a living room with a reading chair beside a window, warm evening light and room for a product on the side table.
The model offers 1K, 2K and 4K resolution choices. Choose the setting that fits the stage of your project, and review the credit quote before generating. Inspect the result before adding more instructions: identify the change that will make the biggest difference first.
Point to the Change in Magic Edit
Open your image in Magic Edit and use Point & Edit to point to an area or draw a region. Write a note describing the intended change. For example: “Replace this chair with a compact blue armchair, keeping its position beside the window.”
Create another note for a separate request, such as changing a lamp or adjusting an object on the table. Keeping the requests separate makes your intention easier to understand and your result easier to evaluate.
Attach a Reference to the Relevant Note
If you have a specific chair or product in mind, attach its reference image to the note that describes it. A reference linked to the chair request has a different role from one linked to a product request. Give each note the visual information it needs.
Use global instructions for direction that applies to the entire image, such as the overall lighting or visual treatment. Keep local notes focused on their own changes. The canvas supports up to eight notes and nine additional references; the selected model's lower limit takes precedence. FLUX 3 Image editing accepts up to ten input images including the source.
Generate, Compare and Refine
Choose FLUX 3 Image editing in the model picker and check the displayed settings and credit amount. Magic Edit supports compatible editing models, so verify your selection before generating.
Review the new image against each note. For furniture, check legs, perspective and shadows. For a product, inspect its shape, text, edges and contact with the surface. Also look outside the requested area for changes you did not intend.
Pointing to a region communicates your intent; it does not guarantee exact geometry or that every other pixel stays untouched. Keep successful versions and make a focused next change rather than piling several uncertain requests onto one attempt.
Explore other image workflows in the Nano Banana 2.1 introduction and the Ideogram 4.5 launch.
Frequently Asked Questions
Is FLUX 3 Image the same as FLUX 3 video?
No. They are separate model families for different output types. Select the image model for image creation or editing.
What resolutions can I choose?
FLUX 3 Image offers 1K, 2K and 4K. Use the settings and credit quote shown in the app for your current task.
How do references work in Point & Edit?
Attach a reference to the note it should guide. The canvas supports nine additional references, subject to the selected model's limit; FLUX 3 Image editing allows ten input images including the source.
Does marking a region guarantee an exact edit?
No. Region notes guide the requested change. Review the whole output, including boundaries and areas you intended to preserve, before using it.