How to Use FLUX 3 Max Image to Video (With Sound)

FLUX 3 Max image to video takes one reference image and uses it as the first frame of your clip. Your prompt then drives the motion and the sound, which the model generates in the same pass. On VIDEO AI ME, pick FLUX 3 Max, attach one image, choose duration, resolution and ratio, and write motion, not appearance.

This is the mode to reach for when the thing on screen has to be yours: a product shot from your last photo session, or a client's approved key visual. You keep the look you paid for and spend the prompt on what the photo cannot supply, which is movement, camera and sound.

What does FLUX 3 Max image to video actually do?

FLUX 3 Max is a video model from Black Forest Labs, the team behind the FLUX image models. 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. On VIDEO AI ME it sits in the editor's video model picker as FLUX 3 Max.

The mode is chosen by what you attach:

Three things follow from "your image is the first frame":

• The opening is decided before you write a word. Composition, framing and lighting come from the image. A photo framed badly for the placement gives a clip that starts badly framed.
• Everything after frame one is generated. Each later frame is new, which is why you check the product in the middle and at the end, not just at the start.
• Sound is still generated. Attaching an image changes where the picture starts, not whether audio is produced.

Step 1: Prepare the image for the frame you will deliver

Match the image to the placement. With the aspect ratio on auto, FLUX 3 Max follows the shape of your image. A vertical photo gives a vertical clip, a square photo a square one. If you pick a named ratio instead, that ratio overrides the image's shape. The simplest way to avoid surprises is to crop the photo to the ratio you plan to publish before you attach it. For Reels, Meta's own Reels ads guidance points to vertical 9:16 video, and TikTok's in-feed video ad specs list 9:16 as the recommended ratio.

Leave room for motion. A product that fills the whole frame leaves the camera nowhere to go. Give it space if you plan a push-in, an orbit or a hand entering the frame.

Use directional light and the original file. A key light from one side creates shadows that can shift as the camera moves. Use the original export, not a compressed copy from your product page.

Do not rely on small text staying perfect. Your label is exact in frame one, because frame one is your photo. In generated frames, small on-screen text and exact logos are unreliable on every video model, so add your logo and any text as overlays after export, in your own editing tool.

Step 2: Pick FLUX 3 Max, attach one image and set the generation

Open a project, choose FLUX 3 Max in the video model picker and attach one reference image. That attachment is what switches the generation to image to video.

Then set the three options:

One generation runs up to 20 seconds with no stitching. Twenty seconds is room for one beat done well, not a montage, so if your ad needs three scenes, plan three clips and join them in the editor.

Step 3: Write motion and sound, not appearance

The image already carries the look. A prompt that re-describes the product, the colours and the background spends words on things the model can already see, and it can pull against the reference. Write about change over time instead.

The method is one paragraph per shot, in a fixed order. For image to video the first part becomes "what moves" rather than "who is there":

1. Subject and action. What moves and how. One focused beat.
2. Camera. Framing and movement: "slow push-in", "handheld, eye level", "slow half-orbit".
3. Light and look. Only if it changes during the shot, such as light sweeping across the product. Otherwise the image has it covered.
4. Sound. Spoken lines in double quotes with who says them, then ambience and effects, then music or "no music".

Here is the shape of it, for a product photo of a bottle on a marble counter:

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.

Notice what is missing: no description of the bottle, the marble or the light. The image supplies all of that. A prompt like "a premium glass bottle with a blue label on a white marble counter" describes a still, gives the model nothing to animate, and restates what the image already fixed.

Three habits make the sound part land:

• Name the sound on the action. "The cap clicks as her thumb flips it open" ties the effect to a moment the picture shows.
• Write the ambience even when it feels obvious. "Quiet kitchen ambience" or "soft street noise outside the window". If you leave it open, the model decides for you.
• Budget any dialogue. Roughly two to three spoken words per second, so a 10 second clip holds about 20 to 25 words. Write fewer than you think.

The full four-part method, with more examples, is in the FLUX 3 Max prompt guide, and the audio side gets its own treatment in FLUX 3 Max sound and dialogue prompts.

Step 4: Draft at 720p, then deliver at 1080p

A 10 second clip costs 170 credits at 720p and 290 at 1080p, so drafting at 720p makes the clips you throw away cheaper. Review each draft against a short checklist:

• Does the motion match what you wrote, or did the model add movement you did not ask for?
• Does the product still read correctly in the middle and at the end of the clip, not only in frame one?
• Did the sound land on the action you described?
• If there is a spoken line, is it the line you wrote, and does the lip sync hold up? Watch the generated speech and lip sync before any clip goes live.

Prompt adherence is the developers' headline claim for FLUX 3 Max. This checklist is how you check that claim on your own product rather than taking it on trust.

When a draft passes, run the same prompt, image and ratio at 1080p for delivery. The 1080p render is a new generation, not an upscale of your draft, so review it with the same checklist before it ships.

Step 5: Finish the clip in the editor

The output is an MP4 in your project, ready for the editor, captions and export. Add captions for viewers who watch with the sound off and join extra scenes on the timeline, then add your logo and any other on-screen text as overlays after export. Audio is always generated with the clip, so if you want your own voiceover or a licensed track instead, replace the track with your own audio afterwards.

Which shots work well from a single image?

Each of these is one beat, which suits a single generation of 5 to 20 seconds.

For the last one, keep the line short, write who says it, and check the lip sync closely before you publish. Ready-made prompts for the product shots are in FLUX 3 Max product ad prompts.

What does FLUX 3 Max image to video not do?

How much does FLUX 3 Max image to video cost?

Image to video and text to video cost the same. The rate is 17 credits per second at 720p and 29 credits per second at 1080p, with audio included in the same generation.

On plan terms, a month's credits cover this many 10 second clips:

The useful number is cost per approved clip, so budget two or three 720p drafts per final. Plan details are on the pricing page, and the wider cost picture, including how FLUX 3 Max compares with other models, is in FLUX 3 Max pricing.

When should you use text to video instead?

Use image to video when the picture has to be exact at the start: your product, your packaging, your approved visual. Use text to video when you need a person, a room or a scene that does not exist yet, such as a creator-style testimonial. That workflow is in how to use FLUX 3 Max text to video. Many product campaigns use both, joined in the editor: a text to video hook, then an image to video clip of the real product.

Frequently Asked Questions

How does FLUX 3 Max image to video work?

Attach one reference image in the editor and FLUX 3 Max uses it as the first frame of the clip. Your prompt describes the motion, the camera and the sound, and the model generates the rest of the frames and the audio in the same pass, from 5 to 20 seconds.

Can FLUX 3 Max use more than one reference image?

No. On VIDEO AI ME, FLUX 3 Max takes one reference image, used as the start frame. If a shot needs a person, a product and a location from separate images, Seedance 2.5 and MiniMax H3 Max accept several reference images.

What aspect ratio does FLUX 3 Max use with an image attached?

On auto, the clip follows the shape of your image. A named ratio such as 9:16 or 16:9 overrides that. The safest habit is to crop the image to the ratio you plan to publish, then leave the setting on auto.

Does FLUX 3 Max image to video generate sound?

Yes. Audio is always on in both modes. Every clip comes back with ambience, effects, music and any spoken lines you write into the prompt. To use a different track, replace the audio with your own audio afterwards.

Can I set an end frame in FLUX 3 Max?

Not in the editor, which has no end-frame control. The VIDEO AI ME API and the make_video agent skill accept FLUX 3 Max with an optional end image. In the editor, you control the first frame through the image you attach.

How much does FLUX 3 Max image to video cost?

The same as text to video: 17 credits per second at 720p and 29 credits per second at 1080p. A 10 second clip is 170 credits at 720p or 290 credits at 1080p, with audio included.