Flux 3 Ocean Dynamics
An official Flux 3 launch sample showing storm-driven waves, spray, and seabirds along a coastal scene.
Flux 3 AI Video and Image Model
Flux 3 AI Video and Image Model
Flux 3 is Black Forest Labs' multimodal model for video, images, native audio, and action prediction. Explore confirmed Flux 3 features and early-access plans.
- Native Video with Synchronized Audio: Flux 3 can generate diverse videos with native audio in a single pass, with clips up to 20 seconds during early access. Black Forest Labs highlights synchronized speech, sound effects, and physical events, but the published evaluations remain preliminary.
- Multimodal Inputs and References: Flux 3 supports text-to-video, image-to-video, video-to-video, and video-audio continuation. Images can act as starting frames or visual references, while source clips can carry a character or central element into a new scene.
- Keyframes, Longer Sequences, and Typography: Flux 3 includes keyframe-to-video transitions, multilingual dialogue, animated typography, varied aspect ratios, and agentic chaining of clips. Black Forest Labs says visual references can help preserve characters across multi-shot sequences, though independent testing is still limited.
- One Backbone Beyond Video: Flux 3 also targets image synthesis and editing plus action prediction from the same multimodal backbone. Image early access is planned after the video preview, while action access begins with selected research and commercial partners such as mimic robotics.
- A Planned Open-Weight Multimodal Model: Flux 3 Dev is planned as an open-weight backbone for video, audio, images, and action prediction. Black Forest Labs has not yet published the release date, license, model size, hardware requirements, API pricing, or final production benchmarks.
- Define the Scene and References: Start with the subject, environment, action, camera movement, duration, aspect ratio, dialogue, and sound. When Flux 3 access supports references, add a starting image, keyframe, or source clip and identify the details that must remain consistent.
Black Forest Labs Multimodal ModelExplore AI Video
Flux 3 AI Video and Image Model
Flux 3 is Black Forest Labs' multimodal model for video, images, native audio, and action prediction. Explore confirmed Flux 3 features and early-access plans.
Official Flux 3 Video Showcase
Watch four launch samples published by Black Forest Labs on July 23, 2026. These official Flux 3 clips show coastal weather, first-person racing, a mirrored dance sequence, and warm clay animation.
What Makes Flux 3 Different
01 / Video • Audio
Native Video with Synchronized Audio
Flux 3 can generate diverse videos with native audio in a single pass, with clips up to 20 seconds during early access. Black Forest Labs highlights synchronized speech, sound effects, and physical events, but the published evaluations remain preliminary.
02 / Text • Image • Video
Multimodal Inputs and References
Flux 3 supports text-to-video, image-to-video, video-to-video, and video-audio continuation. Images can act as starting frames or visual references, while source clips can carry a character or central element into a new scene.
03 / Control • Continuity
Keyframes, Longer Sequences, and Typography
Flux 3 includes keyframe-to-video transitions, multilingual dialogue, animated typography, varied aspect ratios, and agentic chaining of clips. Black Forest Labs says visual references can help preserve characters across multi-shot sequences, though independent testing is still limited.
04 / Image • Action
One Backbone Beyond Video
Flux 3 also targets image synthesis and editing plus action prediction from the same multimodal backbone. Image early access is planned after the video preview, while action access begins with selected research and commercial partners such as mimic robotics.
05 / Dev • Open Weights
A Planned Open-Weight Multimodal Model
Flux 3 Dev is planned as an open-weight backbone for video, audio, images, and action prediction. Black Forest Labs has not yet published the release date, license, model size, hardware requirements, API pricing, or final production benchmarks.
How Flux 3 Fits a Video Production Workflow
This workflow is based on capabilities announced by Black Forest Labs and is intended for approved early-access users. Features, interfaces, and usage terms may change before general availability.
Define the Scene and References
Start with the subject, environment, action, camera movement, duration, aspect ratio, dialogue, and sound. When Flux 3 access supports references, add a starting image, keyframe, or source clip and identify the details that must remain consistent.
Generate and Evaluate the Core Shot
Use Flux 3 to create the main shot, then review anatomy, object persistence, lip sync, audio timing, physical motion, typography, and camera continuity. Change one variable at a time instead of rewriting the entire prompt after every preview.
Extend, Chain, and Finish
Black Forest Labs says visual references and keyframes can help chain clips into longer, multi-shot sequences. Production use and public sharing remain subject to the applicable Flux 3 access terms and permissions.