01 / FLUX 3 MAX
A world in motion
A wide moving scene makes camera direction and changing perspective easy to inspect. For your own take, specify what the camera follows, what remains in frame and how the shot should finish.
Create 5–20 second videos from text or images, with audio and 720p or 1080p output.
0 / 10,000
One video per request. Confirmed failed jobs return credits. View packages
Official FLUX 3 Max excerpt · fal
A wide moving scene makes camera direction and changing perspective easy to inspect. For your own take, specify what the camera follows, what remains in frame and how the shot should finish.
Suggested prompt · written for this studio
A sleek futuristic train rushes past an old frontier station, lifting a cloud of dust. A row of fictional adult travelers watch from a timber platform. Keep the camera steady, with warm daylight and aged film texture. Sound: a rising hum, a fast whoosh and startled murmurs.
Official fal samples are shown for reference. Suggested prompts are not the original sample inputs.
FROM THE FLUX 3 FAMILY
FLUX 3 Max is a video generation model post-trained by fal Research in collaboration with Black Forest Labs. It builds on FLUX 3 with work on visual fidelity, motion and inference speed. Here, you can turn a written scene into a video, or animate an uploaded image with an optional ending frame.
The useful question is what the camera should see next. A product can turn toward the light; a character can glance across a room; a landscape can open up as the camera rises. Describe the action and its sound, choose the length and framing, then review a single generated take. Keep the first test short before spending credits on a longer sequence.
Looking for still-image generation instead? Open the FLUX 3 Image generator for reference-based image editing and 1K–4K stills.
SEE THE MOTION
Watch a wide moving scene, a restrained close-up and a playful science-fiction idea. Inspect the movement and composition before adapting a prompt for your own shot.
01 / FLUX 3 MAX
A wide moving scene makes camera direction and changing perspective easy to inspect. For your own take, specify what the camera follows, what remains in frame and how the shot should finish.
02 / FLUX 3 MAX
Use a close-up to explore an eyeline, a small gesture or a change of expression. Give the subject one clear action and keep the camera instruction restrained; several competing movements make the result harder to evaluate.
03 / FLUX 3 MAX
An unusual combination can make a scene memorable. Anchor the environment first, then describe the interaction between subjects so the idea has a visible beginning, action and end.
These videos were published by fal as FLUX 3 Max examples. They are not generations from this website. The prompts below are new suggestions, not the original inputs. View the official source ↗
01 / TEXT → MOTION
FLUX 3 Max text to video starts with your description of the scene. Put the subject first, then the action, camera and atmosphere. A slow push-in and a fast orbit are different instructions: choose the one that carries the moment. For the first generation, avoid packing an entire story into five seconds. Review whether the main action reads clearly before adding a second beat.

FLUX 3 Max image to video takes an uploaded PNG, JPEG or WebP as its starting frame. A product photograph, a character image or a concept still gives the model a visual starting point. Describe how the scene changes rather than repeating everything already visible. You can add a second image as the desired ending frame; a transition still needs enough time to unfold.
Need a starting image? Create one in the image studio, download it, then upload it as your first frame.
Switch native audio on to request sound alongside the video. Include the important cues in the prompt: a bicycle bell, a soft room tone, a line of dialogue or rain on a roof. Switch it off when you plan to add your own music, voice-over or sound design. Audio is optional, and it does not change this studio’s credit quote.
For a broader look at the model family, read our introduction to FLUX 3.
THE CONTROLS
Choose a format for the intended placement, then adjust length and resolution together. Every selection in the studio is passed to the actual Max endpoint.
5 to 20 seconds, in one-second steps. Start with a short clip to test the motion before committing to a longer take.
720p or 1080p. Higher resolution costs more credits; it does not make a confused prompt more specific.
Auto, 21:9, 2:1, 16:9, 4:3, 1:1, 3:4 or 9:16. Use vertical framing for a phone-first composition and landscape for a wider scene.
High or low. High gives the model more effort to interpret the prompt and can take longer; low is an option for simpler instructions.
Image mode accepts one start frame and an optional end frame. This studio accepts JPEG, PNG and WebP files up to 10 MB each.
On by default, with an explicit switch for silent output. Write sound cues alongside the visual action rather than relying on a separate negative prompt.
A SIMPLE WORKFLOW
Use text to video for a new scene. Choose image to video when you already have a first frame, and add an ending frame only when you want to guide where the clip lands.
Name the subject, describe one main action and give the camera a clear direction. Add light, texture and sound. Choose a duration that gives the action room to happen.
Sign in, check the quoted credit cost and submit once. Review the result in the studio, then download the MP4. Change one instruction at a time when exploring a new take.
PROMPT WORKSHOP
Use this sequence as a starting point: subject → action → camera → light and texture → sound. The goal is a shot you can imagine watching, with a beginning and an ending. Detailed does not have to mean long.
A ceramic artist lifts a freshly glazed blue cup from a workbench and slowly turns it toward the window. Medium close-up, one gentle camera push, soft morning light and visible glaze texture. Sound: a quiet studio, the cup touching wood and distant birds. One continuous five-second shot, no titles.
Original prompt suggestion, not a claim about a tested output.
Describe movement relative to the uploaded frame: the head turns, fabric catches the wind, or the camera pulls back. Avoid asking for a completely different subject unless a transformation is the point.
Order the actions in time. A character enters, pauses and speaks is easier to follow than three simultaneous directions. Use a longer clip when each beat needs space.
First simplify the action. Then remove competing camera moves and check whether the framing leaves enough room. Change the prompt before increasing resolution or generating a much longer clip.
KNOW THE COST
Credits depend on the selected output resolution and clip length. Text and image modes use the same quote; the native-audio switch does not add a surcharge. The button shows the exact cost before you submit.
Compare packages on the pricing page and choose enough credits for the clips you plan to make.
| Clip length | 720p | 1080p |
|---|---|---|
| 5 seconds | 173 credits | 294 credits |
| 10 seconds | 345 credits | 587 credits |
| 20 seconds | 689 credits | 1174 credits |
These are this website’s credit prices, not fal’s API charges. They use standard provider rates rather than a temporary launch discount. Check your balance before generating; the signup allowance does not cover the minimum Max clip.
CHOOSE THE RIGHT TOOL
They solve different jobs. Use the video model when movement, timing and sound matter. Use the image model when the deliverable is a still, a design or an edited photograph.
For image generation and multi-reference editing, use the dedicated FLUX 3 Image tool. For other video engines, explore the video generator.
Building your own integration? See the official fal API documentation for schemas, access and current upstream pricing.
FLUX 3 Max creates video clips from written prompts or images, with optional generated audio. It is useful for exploring a scene, a visual transition, a character action or a short concept. It is a video model; use FLUX 3 Image when you need a still image.
fal Research post-trained FLUX 3 Max in collaboration with Black Forest Labs, using FLUX 3 as the foundation. This website is an independent interface and uses fal’s Max endpoints for generation.
You can watch the official examples and explore prompts without paying. Generating a video requires sufficient site credits. The normal signup allowance does not cover the shortest Max generation; see the live quote and pricing table before purchasing.
Choose any whole-second duration from 5 to 20 seconds. A short clip is a good way to check the main action. Longer output costs more credits and gives a sequence more time; it does not guarantee a better result.
This studio offers the provider’s 720p and 1080p output settings. It does not offer 4K video. Resolution labels describe the selected output tier; they are not a promise of a particular result or exact pixel dimensions for every aspect ratio.
Yes. In image-to-video mode, upload a starting image and optionally an ending image. JPEG, PNG and WebP are supported up to 10 MB each. The end frame guides the requested ending; the scene between the two frames is generated.
This studio supports a first frame and an optional last frame. It does not accept a reference video or a list of timed keyframes. fal has a separate keyframes API; its capabilities should not be confused with the controls available here.
Native audio is optional and is enabled by default in this studio. Include the sounds or dialogue you want in the prompt, or disable audio for a silent video that you will edit elsewhere. The audio setting does not change the credit cost here.
fal positions Max for faster video iteration and publishes model benchmarks. Your actual wait also includes queueing, prompt processing, upload and output handling. Treat those benchmarks as provider measurements rather than a guaranteed delivery time for every clip.
fal offers public text-to-video, image-to-video and keyframe endpoints. Developers need their own provider account and API key. This website is a credit-based web studio and does not issue developer API keys.
You can download a completed video as an MP4. That is different from downloading model weights. The service used here is a hosted API; this page does not provide a local model file or a ComfyUI installation package.
A confirmed rejected or failed job returns its generation credits through the site’s refund flow. A network timeout is not proof of failure: the provider may already be processing the request. Do not submit again while the status is uncertain; keep the job ID and contact support if needed.
Start with one clear moment. Give it an action, a camera and a sound.
Open FLUX 3 Max