Now loading...
Beginning today, the initial version of FLUX 3 Video is officially accessible through the BFL API and select partners. This advanced model can create video clips of up to 20 seconds in length with HD resolution, and Full HD output is attainable through upscaling, while native audio accompanies the video.
FLUX 3 represents a sophisticated multimodal model capable of generating and predicting video, audio, images, and actions. This marks the first broad release of its video generation features. For further details on FLUX 3, visit here.
Designing a comprehensive representation of reality necessitates a multimodal approach, as each individual format—such as images or videos—captures only a fraction of it. Recognizing this, FLUX 3 is crafted to represent reality as accurately as possible, avoiding confinement to any singular style. The resulting video outputs span a wide range, offering aesthetics that can be raw, playful, or even surreal.
This versatility makes FLUX 3 a powerful tool for content creators. The model’s capabilities allow it to manage both simpler and intricate prompts while deeply understanding contextual nuances. It can fluidly transition between scenes and camera perspectives within a single generation, incorporate text seamlessly, and produce dialogue across multiple languages with realistic accents and synchronized lip movements. Users can achieve results that are personal, stylized, or purely entertaining.
Available to the general public, FLUX 3 Video can generate video clips with a maximum length of 20 seconds and native audio in resolutions of both HD (720p) and Full HD (1080p). Key features include:
Text-to-Video: Users can describe scenes in simple language or provide detailed prompts, with the model creating natural movements, scene coherence, and sound. Image-to-Video and Keyframes: Starting with a single image, users can define endpoints or keyframes, allowing sequential video creation that adheres to the desired visual style. Video Continuation: FLUX 3 can extend existing video and audio clips by considering up to four seconds of provided content and determining subsequent actions. Creating Multiple Shots: Multiple scenes and camera angles can be seamlessly integrated within one video while maintaining narrative coherence. Audio and Dialogue: The model can generate dialogue, sound effects, and ambient noises aligned with each frame. Multi-Linguality: It supports a high number of languages including English, Spanish, and Chinese, providing precise synchronization for lip movements. World Knowledge and Grounding: Combining pretrained knowledge with real-time contextual grounding, FLUX 3 serves well for educational content. Draft Mode: This feature allows users to preview ideas rapidly at lower costs, enabling iterative refinement before committing to high-quality outputs.
FLUX 3 Video offers state-of-the-art capabilities for both text-to-video and image-to-video generation. Evaluations against leading models have shown it to surpass them in performance, particularly noted in human assessments of its text and image generation abilities since its initial announcement. Internal tests have demonstrated FLUX 3 as the preferred choice for these tasks.
The development of FLUX 3 Video is approached with responsibility, involving multiple safeguards designed to prevent misuse. Collaborating with trusted third-party organizations like Cinder, FLUX 3 has been scrutinized for various risks prior to its public release.
Looking ahead, future updates will introduce greater control features and expand generation capabilities by integrating various media types, including images and audio. Additionally, forthcoming releases will include FLUX 3 Image for image generation and editing, as well as FLUX 3 Dev, an open-weight variant.
FLUX 3 Video is currently available via the BFL API and through selected partners, and anticipation is high for the creative projects that users will undertake with this new tool.
For more information about FLUX 3, visit here. For documentation, click here.
