Generation video FLUX 3 Video

FLUX 3 Video is the video generation model from Black Forest Labs, the lab behind the FLUX image family. Its defining feature is timeline keyframes: instead of accepting a single starting image, it lets you pin up to ten still images along the clip's timeline and animates the motion between them. One image sets the opening frame, two set the first and last frames, and three to ten distribute those plus evenly spaced beats in between — so you can storyboard a shot as stills and have the model fill in the motion.

The keyframes are timeline anchors, not identity references. FLUX 3 does not treat them as character or style reference images the way reference-to-video models do; each pinned still is a moment the video passes through. If you need a character to stay on-model across free camera movement, a reference-driven model such as Seedance 2.5 or Kling V3 Omni is the better tool. If you know exactly which frames the shot must hit, FLUX 3's keyframe timeline is the more direct control.

Audio is native and synchronized, including lip-sync for on-screen dialogue, and costs nothing extra. Clips run 5 to 20 seconds in whole-second steps, and the aspect-ratio options include a 21:9 cinematic format alongside the usual 16:9, 9:16, 1:1, 4:3 and 3:4 (or adaptive, following the first keyframe).

Three quality tiers are available. HD is the standard tier at 11.3 credits per second — 57 credits for a 5-second clip, 113 for 10 seconds. Draft renders at roughly a third of the HD rate and is the right tier for iterating on keyframe placement before a final render. FHD is an upscale of the HD render at about 1.7x the HD rate; it sharpens the output rather than re-rendering at higher native detail. The exact total is always shown before generation, and no separate Black Forest Labs account is required.