How to Use FLUX 3: Photos, Text in Images and Video in 2026

To use FLUX 3, open it in the IIshki studio, write your prompt, pick an aspect ratio (from 21:9 to 9:16) and a resolution from 768 to 4K, and the Black Forest Labs model returns a photorealistic image with accurate text. To edit an existing photo or assemble a scene, upload up to 10 references and say what to take from each. If you need a clip, FLUX 3 Video sits right next to it: 5 to 20 seconds of video with sound and speech, up to 4K. Both models work without a VPN or a foreign card. Below: how to write prompts, ready-made examples and the most common mistakes.

What FLUX 3 is and what it can do

FLUX 3 is the new generation of models from Black Forest Labs, the team behind FLUX.1 and FLUX.2. In our studio it comes as two separate cards:

• FLUX 3 for images. Text-to-image, photo edits by instruction ("change the jacket color", "remove the sign"), style transfer and scenes built from several references.
• FLUX 3 Video for video. A clip from text, from storyboard frames, or a continuation of an uploaded video; ambient sound and lip-synced speech are generated together with the picture.

The image model is strongest at skin and material texture, photographic lighting and readable lettering. That is why people pick FLUX 3 for portraits, product shots, posters and covers where text has to be part of the scene, not a sticker on top.

Which model fits the task

To compare FLUX 3 with other image generators, see our roundup of the best AI image generators; for video models, see the best AI video generators.

How to use FLUX 3 in 5 steps

1

1. Open the model and pick a format

Go to FLUX 3. Aspect ratio offers Auto and eight formats: 1:1, 16:9, 9:16, 21:9, 4:3, 3:4, 3:2 and 2:3. Use 9:16 for Stories and Reels, 16:9 for a YouTube thumbnail, 3:4 or 1:1 for a product card. In Auto mode with a reference, the frame follows the proportions of the first uploaded image.

2

2. Describe the picture in order

FLUX 3 understands a prompt best when it reads like a description of a finished shot: the main thing first, details after. A handy order is type of shot → subject → place → light → angle and lens.

Studio portrait of an elderly fisherman in a knitted sweater, sitting by the window of a wooden house, soft side daylight, natural skin texture, 85 mm lens, shallow depth of field

Write what should be in the frame, not what should be missing. Instead of "no people", write "an empty street at dawn". Words like "masterpiece" and "8K" change nothing except prompt length: resolution has its own switch.

3

3. Add text if you need it

Put the words in quotes and say right away where they sit and how they look.

Coffee shop poster, in the top third the large words "MORNING STARTS HERE" in a bold cream-colored retro typeface, below it a cappuccino on a wooden counter, warm morning light

One or two short lines render more reliably than a paragraph. For several captions, list each one separately with its own place in the frame.

4

4. Upload references to edit or build a scene

The Image slot takes up to 10 JPG, PNG or WebP files (at least 256 px per side, up to 30 MB). Refer to them in upload order and give each a role: character, object, background, style.

Take the woman from image 1 and seat her at a table in the interior from image 2, clothes as in image 3, light and color as in image 2

For editing a single photo, use a three-part formula: what changes → how it changes → what stays the same.

Replace the red jacket with a dark navy coat. Keep the face, hair, pose, background and lighting unchanged

One edit at a time gives predictable results. If you need another change, run the next edit on the new result and list what to keep again.

5

5. Choose a resolution and run it

Resolution steps are 768, 1K, 1.5K, 2K (the default) and 4K. A working routine: tune the prompt at 768 or 1K, and once the composition is right, render the final at 2K or 4K. Keep in mind that a new resolution is a new generation, so details may shift; if the exact frame matters, generate it at the final size from the start. The price of each step is on the Generate button. For heavy upscaling of a finished image there is a separate Topaz upscaler.

How to make a video in FLUX 3 Video

FLUX 3 Video has two modes:

• Frames A clip from text or from pictures. The first picture becomes the opening frame, the last one the closing frame, and the rest (up to 10 in total) fall in between.
• Video Upload a clip of up to 60 seconds and the model continues it.

Length runs from 5 to 20 seconds in one-second steps, quality is 720p (the default), 1080p, 1440p or 4K, and the formats are 16:9, 9:16, 1:1, 4:3, 3:4 and 21:9.

Three rules for a video prompt:

1. Start with the camera. "The camera slowly pulls back", "handheld at eye level": the model does better when motion comes first.
2. Describe the sound. Audio is generated in the same pass as the picture, so "crashing surf, seagulls calling" really changes the result.
3. Put lines in quotes with a speaker. Lips move in time with the speech.

Handheld camera at eye level. A lighthouse keeper in a storm looks into the lens and says: "The storm is coming". Wind whips his coat, surf and howling wind on the soundtrack

Test the idea first with a 5-second 720p clip, the most affordable option. When the motion and sound work, repeat at the length and quality you need. You can make the first frame in FLUX 3 and build the whole "image → edit → video" chain on the canvas.

Common mistakes

FLUX 3 or other models

FLUX 3 is the pick for photorealism and lettering. For stylized illustration and batches of variations, people more often use Seedream 5.0 Pro; our comparison shows how it stacks up against GPT Image 2. If you are looking to replace Midjourney, see Midjourney alternatives. Ready-made workflows with FLUX 3 are in our guides on product backgrounds and making a logo with AI, where FLUX 3 puts a logo on a mockup and places a product into a scene.

What's next

Start with a single portrait or poster in FLUX 3: describe the shot in order, put text in quotes, use 768 for the draft. Once the picture is ready, bring it to life in FLUX 3 Video with camera motion, sound and a spoken line. The price of every generation is shown on the button before you start.