Skip to main content
Turning an image into video is not a single action. Depending on how strictly the model should follow your image — and whether it defines the start, the end, or just the look of a shot — FLUX 3 gives you several distinct ways to animate. The right approach depends on what your image should become: the literal opening frame of the shot, one end of a defined motion, or one of several waypoints the shot moves through. Choosing deliberately gives you far more control than sending the same image into every mode and hoping for the best.

​
Overview

i2v animates from the images you give it — all through one keyframes field. What changes is how many frames you pin and where they sit in the shot:
Reach for Startframe when the result must begin exactly on your image, Start + end frame when you know both ends of the motion, and Keyframes when the shot has to hit several specific moments along the way. Timestamped and 3+ keyframe requests need an explicit duration.

​
Startframe

In i2v, your image sets the opening frame and the prompt drives the motion. Press and hold any frame below to preview the video FLUX 3 generates from that single still — release to snap back to the source image.

​
Examples

Notice how the source frame is preserved exactly at the start of each clip — that first-frame fidelity is what makes Startframe the most literal way to animate an image.

​
Start + end frame

Give FLUX 3 two stills — the frame the shot should open on and the frame it should land on — and it generates the motion that carries one into the other. Reach for this when both ends of a shot matter: a reveal that has to resolve on a specific composition, a transformation with a defined before and after, or a loop that must return to where it began. Keep the two frames related — same subject, scene, or camera setup — so the interpolation has a plausible path between them. The wider the gap between start and end, the more freedom (and unpredictability) you hand the model.

​
Examples

Each example shows the two input frames — the start and the end — followed by the clip FLUX 3 generated to bridge them.
Start frame — city skyline by day

Start frame

End frame — city skyline at night

End frame

​
Keyframes

Pass several stills as ordered waypoints and FLUX 3 interpolates a single continuous shot that moves through each one in turn. Where Start + end frame pins only the two ends, keyframes let you choreograph the moments in between — hit this pose, then this composition, then this one — while the model handles the motion connecting them. Order matters: the frames are read as a timeline from first to last. Space them so each transition is achievable, and keep the look consistent across frames so the shot reads as one take rather than a series of cuts.

​
Examples

Each example shows the ordered keyframes pinned along a 5-second clip ([[0, …], [2.5, …], [5, …]]) followed by the shot FLUX 3 interpolated through them.
Keyframe 1 — magenta aurora

Keyframe 1 — 0s

Keyframe 2 — teal aurora

Keyframe 2 — 2.5s

Keyframe 3 — green aurora

Keyframe 3 — 5s

Video Prompting Overview

Workflow selection and the broader FLUX 3 video prompting framework.

Text-to-Video

Generate a shot from a text prompt when you don’t need a fixed first frame.

Camera Terms

Reusable framing, angle, movement, and focus phrasing for your prompts.

FLUX 3 Video

Modes, examples, and workflow breakdowns for FLUX 3 video.