FLUX 3 Video

FLUX 3 Video generates clips with synchronized audio from text, pinned images or an existing video. Image mode accepts up to ten timed keyframes. A separate editing endpoint changes an existing clip.

Preview

Model details

Draft mode produces an HD preview and a cache for a full-quality render. Higher-resolution outputs use the same generation followed by a video upsampler. The documented latest version can change as BFL updates the preview.FLUX 3 Video official documentation

Release and access details

The release notes label this model a preview. Text and image generation allow up to 20 seconds. Video continuation is limited to 15 seconds.FLUX release notes

Practical use

Practical guidance. Plan the clip around a small number of actions that fit its duration. Check sound, dialogue timing and motion together. Keep the draft cache when selecting a take, since rerunning a prompt is a different operation from enhancing the chosen draft.

Specifications & access

FLUX 3 Video specifications
DetailValue and conditions
StatusPreview
AccessAPI, Provider product
Model identifiersflux-3-video
Access termsProprietary hosted service; no public weights verified.FLUX 3 Video official documentation
AudioGenerated by default; can be disabled.FLUX 3 Video official documentation
Delivered UHD at 16:93840 × 2176 pixels, with an upsampling stage.FLUX 3 Video official documentation

Related models

  • Video super-resolution model with precise and creative modes that retains the original audio.

    Launch date not confirmedAPI
  • FLUX 3 video model post-trained by fal Research with Black Forest Labs for text, image and keyframe inputs.

    Launch date not confirmedAPI · Provider product
  • Text-guided video editing model for object replacement, style changes, scene changes and dialogue edits.

    Launch date not confirmedAPI