FLUX 3 Max is fal Research’s post-trained video model, developed with Black Forest Labs on FLUX 3. Create a scene from words or a starting image, guide its final frame, and bring movement and sound into one creative brief.
Begin with an action the viewer can recognize. The text route creates a new scene without a starting image: name the subject, setting, camera movement, and sound. A dancer releasing a silk ribbon gives the shot a visible event to follow. Decide where that movement begins and how it settles before adding atmospheric details. A product video can use the same subject–action–setting structure.
Use a photograph or illustration to establish the opening composition, then describe the movement that follows. The fal image route accepts PNG, JPEG, and WebP. A puppy poised on a lavender path can take its first steps as the camera moves beside it; the source image supplies appearance while the prompt directs the next moment.
An optional ending image gives image-to-video a second visual milestone. Establish the closed doors in the opening frame and the ocean terrace in the ending frame, then describe the action connecting them. Keep subject identity, perspective, and the intended transition coherent across both supplied pictures.
Audio generation is enabled by default and can be switched off. Describe the audible event as clearly as the visible one: piano notes, a spoken sentence, or waves against a shore. A performer at a grand piano connects the sound to a visible source. Listen to the result alongside the movement when choosing a take.
Write the camera’s relationship to the action: follow the horses from the side, move toward a face, or hold a wide composition. The API offers low and high reasoning effort, with high as the default; greater effort takes longer. This gives a creative brief both a visual direction and a deliberate processing choice. Shape the movement around the purpose of your cinematic video.
The reviewed fal routes accept 5–20 seconds in one-second steps. Start at 720p or select the documented 1080p upscale option. Text prompts can use landscape, portrait, square, and wider compositions; automatic framing uses 16:9 for text and follows the starting image for image-led work. Plan the frame around the intended delivery.
Compare creative inputs, shot controls, sound, and clip length across the provider workflows.
| Dimension | FLUX 3 Max | FLUX 3 | Wan 3.0 |
|---|---|---|---|
| Model role | fal Research post-training on FLUX 3 | Black Forest Labs foundation video model | Alibaba all-in-one generation and editing |
| Creative starting points | Text; starting image with optional ending image | Text, image, first/last frames, keyframes, existing clip | Text, frames, multimodal references, file or web context |
| Clip length | 5–20 seconds on the reviewed fal routes | Up to 20 seconds | 2–30 seconds without reference video |
| Sound | Generated audio; enabled by default | Native audio | Native dialogue, music, and sound effects |
| Workflow controls | Motion prompt, ending frame, reasoning effort | Keyframe, extension, draft, and enhancement routes | Reference, editing, and extension tasks |
Plan one reveal: a hand lifts a perfume bottle, light catches the glass, and the camera moves closer. A starting image can establish the arrangement before you direct the action and its timing.
Describe the gesture, the camera’s distance, and the sound around the performer. A short continuous movement gives expression and body motion a clear role in the finished scene.
Begin with an illustrated character and supply a final composition when the destination matters. Let the prompt explain the transition, rather than asking two unrelated frames to connect without a story.
Explore Pexo Create workflows for product messages, documentary moments, and animated stories.
FLUX 3 Max is a video model post-trained by fal Research in collaboration with Black Forest Labs, using FLUX 3 as its foundation. The fal workflows reviewed here generate video from a text prompt or a starting image and can include generated audio.
Yes. The image-to-video API requires a starting image and accepts an optional ending image. PNG, JPEG, and WebP are supported. Describe the movement connecting the two compositions in the prompt.
The reviewed fal text and image routes support 5–20 seconds, with five seconds as the default. Output defaults to 720p, with a documented 1080p upscale option. Automatic aspect ratio uses 16:9 for text prompts and follows the starting image otherwise.
The API offers low and high reasoning effort, with high selected by default. Higher effort increases generation time. The seed is optional, and audio generation can be enabled or disabled for the request.
FLUX 3 Max is served on fal through its model playground and API. It is the fal Research post-trained variant of FLUX 3; the original Black Forest Labs model has its own documented generation, draft, keyframe, and extension workflows.
FLUX 3 Max is a video model post-trained by fal Research in collaboration with Black Forest Labs, using FLUX 3 as its foundation. The fal workflows reviewed here generate video from a text prompt or a starting image and can include generated audio.
Yes. The image-to-video API requires a starting image and accepts an optional ending image. PNG, JPEG, and WebP are supported. Describe the movement connecting the two compositions in the prompt.
The reviewed fal text and image routes support 5–20 seconds, with five seconds as the default. Output defaults to 720p, with a documented 1080p upscale option. Automatic aspect ratio uses 16:9 for text prompts and follows the starting image otherwise.
The API offers low and high reasoning effort, with high selected by default. Higher effort increases generation time. The seed is optional, and audio generation can be enabled or disabled for the request.
FLUX 3 Max is served on fal through its model playground and API. It is the fal Research post-trained variant of FLUX 3; the original Black Forest Labs model has its own documented generation, draft, keyframe, and extension workflows.
