One brief, picture and sound together
FLUX 3 renders the event, not just the frame
FLUX 3 generates motion and its audio in the same pass, so a useful prompt reads like a compact production brief: what exists, what changes, what causes each sound, and what has to stay true from the first frame to the last.
Build the briefWorld, materials, goal
Time, force, continuity
Cause, distance, rhythm
5–20s, 720p or 1080p
Up to 20 seconds in one generation
Direct time, not a moving thumbnail
Split the clip into readable beats. Establish the world, introduce the force, let the camera respond, and specify the final state. Add sound where it has a cause: edge scrape, wind pressure, breath, an impact, or a deliberate pocket of silence.
- 01Establish
Wide scale, terrain, direction of travel.
- 02Accelerate
Camera pursuit, changing distance, rising sound.
- 03Resolve
Readable final pose, environment still in continuity.
Four starting points
FLUX 3 takes the brief from wherever you are
The same written brief drives every mode. What changes is how much of the shot you have already decided — nothing at all, one frame, both ends, or an approved take that needs to keep going.
Text to video
Nothing but the prompt. Best when the world, camera, and sound are still open and you want the model to propose the whole shot.
Image to video
The uploaded frame fixes subject, palette, and framing. Describe only what changes after it, not what the picture already shows.
First and last frame
Both ends are locked, so the model only has to solve the movement between them. The most controllable mode for a beat you have already storyboarded.
Extend a clip
Continue an approved take past its final frame instead of regenerating it. Motion, scene, and sound carry over; the brief covers what happens next.
Draft first, then full quality
- 01
Run the brief in draft to test timing and framing for a fraction of the credits.
- 02
Fix what the draft exposes: pacing, camera distance, sound placement, final pose.
- 03
Re-run the corrected brief at full quality in 720p or 1080p.
- 04
Extend the approved clip when the shot needs to run longer than one generation.
FLUX 3 prompt structure
Build one world brief
Each layer answers a different production question. Keep them separate so a change to the soundtrack does not accidentally rewrite the camera, character, or final frame.
- 01
World
Name the place, subjects, materials, weather, and physical relationships.
- 02
Time
Describe the action beat by beat, including pace, camera motion, and the ending.
- 03
Sound
Place dialogue, ambience, impacts, music, and silence on the same timeline.
- 04
Continuity
Lock identity, wardrobe, geometry, light direction, and reference roles.
- 05
Delivery
Set duration, aspect ratio, resolution, and the review criteria for the final frame.
A weathered grey horse crosses a windy highland from left to right. Begin wide and low, then track alongside at shoulder height as the pace builds. Hooves compress wet grass; mane and tail respond to gusts. Use distant thunder, layered hoof impacts, breath, and no music. Preserve the horse's markings and travel direction throughout. End on a steady profile frame with open land ahead.
Still searching for “FLUX 3 Image”? It hasn’t been released — see the FLUX.2 still image workflow before you generate the key frame. For accelerated video with native audio, see the SOL-H3 preview.