FLUX 3 Video
Video modelCard summarizedFLUX 3 is our frontier multimodal model for generating and predicting video, audio, images, and actions.
Source: FLUX 3 Video, Part 1: Generation (launch post) and FLUX 3 announcement (no model card published)
Where to call it
- BFL API
flux-3-video
Good for
- Text- and image-to-video with synchronized sound
- Keyframe control and clip continuation
- Multi-shot sequences with continuity across cuts
Overview
FLUX 3 Video
FLUX 3 is our frontier multimodal model for generating and predicting video, audio, images, and actions.
- Max resolution
- 4K uhd
- Max length
- 20 s
- Modes
- text-to-video, image-to-video
- Released
- 4 Aug 2026
FLUX 3 Video is the video part of FLUX 3, Black Forest Labs' multimodal flow matching model trained jointly on images, video and audio. One endpoint generates clips of up to 20 seconds with synchronized audio from text, from pinned keyframes or by continuing an existing clip, with multi-shot scenes, multilingual lip-synced dialogue and output up to 4K via an upsampler. A draft mode gives cheap previews that can be enhanced into the same shot at full quality. Launch post (4 Aug 2026); FLUX 3 post (23 Jul 2026); docs overview
flux-3-video- Typical price
- $0.15 / sec
- Max resolution
- 4K uhd (3840×2176 at 16:9, upsampled from hd)
- Max length
- 20 s (video continuation 15 s)
- Aspect ratios
- 21:9, 2:1, 16:9, 4:3, 1:1, 3:4, 9:16, 9:21
- Frame rate
- 24
- Released
- 4 Aug 2026
- text-to-video
- image-to-video
- first and last frame
- keyframes (up to 10, timed)
- video continuation
- draft and enhance
- multi-shot in one generation
- multilingual speech with lipsync