What Is FLUX 3 Max? The AI Video Model Explained

FLUX 3 Max is a video model from Black Forest Labs, the team behind the FLUX image models. It turns a text prompt, or one reference image used as the first frame, into a 5 to 20 second clip with sound generated in the same pass. On VIDEO AI ME it renders at 720p or 1080p.

That is the short answer to "what is FLUX 3 Max". The longer one matters if you are deciding whether to put ad budget behind it, because the model has a specific shape: what it takes in, what it gives back, what it costs, and the handful of things it does not do. This page covers all four, using only what the model's developers say and what the VIDEO AI ME editor actually exposes.

Who makes FLUX 3 Max, and how does it relate to FLUX 3?

Black Forest Labs built its name on the FLUX image models, and FLUX 3 Max carries that name into video as part of the FLUX 3 family.

FLUX 3 Max is one model within that family. It is a post-trained variant of Black Forest Labs' FLUX 3 Video, which its developers describe as tuned for stronger prompt adherence and better visual quality, and optimized for lower latency. "Post-trained" means the base video model was trained further after its main training run, to shift its behaviour toward specific goals. Here, by the developers' description, those goals are closer prompt following, better visual quality and lower latency.

Keep the framing straight: those three points are the developers' positioning. They describe what the model was tuned for, not a measurement of how it performs on your products. The FLUX 3 Max review guide gives you a short test plan to check them on your own briefs.

What is FLUX 3 Max able to do?

It generates one video clip, with audio, per request. On VIDEO AI ME the mode is picked by what you attach:

Text to video builds the whole shot from your words: subject, action, camera, light and sound.

Image to video starts from a frame you already own, such as a product photo or a campaign still. Because the image already carries the look, the prompt's job is motion and sound. Describing the product's appearance again mostly wastes words.

In both modes, the output is an MP4 that lands in your project, ready for the editor, captions and export.

What does native audio mean here?

Most clips people run on paid social need sound. Plenty of video models return silent footage, which leaves you to source a voice, music and effects, then line them up by hand.

FLUX 3 Max generates the sound in the same pass as the picture. A clip can carry:

• Spoken lines you write into the prompt, in double quotes, with who says them
• Ambience, such as room tone, a street, wind or a busy coffee shop
• Effects tied to the action, such as a cap twist, a pump click or a door
• Music, or a plain "no music" if you would rather add a track later

Placement cues work too: "traffic on the left, a kettle on the right".

Audio is always on for FLUX 3 Max on VIDEO AI ME. Every clip comes back with sound. If you want a different track, replace the audio with your own audio once the clip has landed. And whatever model you use, watch the generated speech and lip sync with sound on before a clip goes live.

How does it read your prompt?

FLUX 3 Max reads plain natural language, and it does one thing worth understanding before you write a word: it reasons about your prompt and expands it before rendering. A short prompt still produces a full shot.

That is useful and a little dangerous. The expansion fills in whatever you leave open. If the product colour, the line of dialogue or the framing matters to the ad, write it. If you leave the ambience out, the model picks one for you. Specific prompts still win.

The prompt structure that works is one paragraph per shot, in this order:

1. Subject and action. Who or what, doing what. One focused beat.
2. Camera. Framing and movement, such as "slow push-in" or "handheld, eye level".
3. Light and look. Time of day, light source, grade.
4. Sound. Spoken lines in quotes with the speaker, then ambience and effects, then music or "no music".

An image to video example, where the image already sets the look:

The bottle slowly rotates on the marble counter as condensation runs down the
glass. The camera makes a slow half-orbit. A crisp cap twist, a soft fizz,
light lo-fi music.

Budget dialogue at roughly two to three spoken words per second of clip, so a 10 second clip holds about 20 to 25 words. The step by step version is in how to use FLUX 3 Max.

What are the FLUX 3 Max specs on VIDEO AI ME?

Every duration up to 20 seconds is one generation, with no stitching. auto renders 16:9 for text to video and follows your image for image to video, and a named ratio overrides it. The 21:9 option suits cinematic placements and brand films. Before you pick a ratio, check the placement's own spec: TikTok's video ad specifications list the vertical, horizontal and square sizes it accepts, and YouTube's upload settings help page gives 16:9 as the standard ratio on a computer.

Agents and developers reach the model too. The make_video skill on the VIDEO AI ME MCP server and /api/v1 accept FLUX 3 Max, and that route also accepts an optional end image. The editor itself has no end-image control.

What can't it do on VIDEO AI ME?

Knowing the edges saves credits. These are the limits, with the model in the same picker that covers each one:

None of these are reasons to skip the model. They tell you which jobs to hand it.

Where does it fit among the other models?

Here is the spec sheet side by side for the models you are most likely to weigh it against:

Read it by job. FLUX 3 Max is built for a single shot of up to 20 seconds, at up to 1080p, with sound, from a prompt or one image. Seedance 2.5 is the pick for longer single generations and multi-reference work; see what Seedance 2.5 is. MiniMax H3 Max covers heavy reference work too, explained in what MiniMax H3 is. Boreal is the one to use when your own voice-over should drive the picture.

How much does FLUX 3 Max cost?

On VIDEO AI ME, FLUX 3 Max costs 17 credits per second at 720p and 29 credits per second at 1080p. One credit is one cent of generation cost, and text to video and image to video cost the same.

For context, FLUX 3 Max at 1080p costs less per second than Seedance 2.5 at 720p, at 48 credits per second. A 20 second clip is 580 credits on FLUX 3 Max at 1080p against 960 on Seedance 2.5 at 720p.

Launch rate: through November 6, 2026, FLUX 3 Max is billed at 5 credits per second at 720p and 8 at 1080p, which makes a 10 second 720p clip 50 credits. The standard rate applies automatically from November 7, 2026.

Plans and allowances are on the pricing page. The full FLUX 3 Max pricing breakdown covers clips per plan and every duration.

What is FLUX 3 Max best for?

It suits buyers who measure cost per usable asset:

• Ecommerce and DTC operators who need product clips with sound and already have product photos to start from.
• Media buyers testing several hooks for a 9:16 or 16:9 placement, where a 10 to 20 second clip with audio is the unit.
• Agencies producing client creative, where prompt adherence, if it holds for your briefs, means fewer rejected drafts per approved asset.
• SaaS and brand marketers who want a 21:9 or 16:9 hero beat at 1080p.

It is a weaker fit if your brief depends on several reference images, on your own recorded audio, or on one continuous shot longer than 20 seconds. Those jobs belong to the models listed above.

Frequently Asked Questions

Who makes FLUX 3 Max?

Black Forest Labs, the team behind the FLUX image models. FLUX 3 Max belongs to the FLUX 3 family, and it is a post-trained variant of Black Forest Labs' FLUX 3 Video. On VIDEO AI ME it appears in the editor's video model picker as FLUX 3 Max.

What is the difference between FLUX 3 and FLUX 3 Max?

FLUX 3 Max is a post-trained variant of Black Forest Labs' FLUX 3 Video, which its developers describe as tuned for stronger prompt adherence and better visual quality, and optimized for lower latency. Those are the developers' claims, so check them on your own briefs with a few short 720p generations.

Is FLUX 3 Max a video model or an image model?

FLUX 3 Max is a video model. Black Forest Labs is best known for the FLUX image models, but FLUX 3 Max generates video clips with sound. On VIDEO AI ME it can start from a text prompt or from one image used as the first frame.

Does FLUX 3 Max generate sound?

Yes. FLUX 3 Max generates audio in the same pass as the picture: spoken lines you write into the prompt, ambience, effects and music. On VIDEO AI ME audio is always generated, and you can replace the track with your own audio afterwards.

How long can a FLUX 3 Max video be?

Up to 20 seconds in a single generation. The VIDEO AI ME editor offers 5, 8, 10, 12, 15 and 20 seconds. For a concept with several scenes, generate each scene as its own clip and join them in the editor.

Can FLUX 3 Max use reference images?

It takes one reference image, used as the first frame of the clip. FLUX 3 Max is not a multi-reference model on VIDEO AI ME. If you need several reference images to hold a character or product consistent, Seedance 2.5 and MiniMax H3 Max take several.