Black Forest Labs has released Flux 3, a multimodal foundation model capable of generating videos with native audio for the first time, supporting clips up to 20 seconds long.
Flux 3 represents a significant step forward in AI video generation by integrating audio directly into the creation process. The model learns from images, video, and audio data, enabling it to produce synchronized video and sound without post-processing.
Black Forest Labs conducted internal testing that placed Flux 3 ahead of Seedance 2.0, the current market leader in video generation. However, independent verification of these results is not yet available.
The 20-second duration marks an expansion in video length capabilities for the company. This native audio integration eliminates the need for separate audio generation or synthesis tools, streamlining the video creation workflow.
Beyond video generation, Black Forest Labs is exploring broader applications for Flux 3. The company is actively testing the model on robotics tasks, suggesting ambitions to develop what it describes as a world model—a system capable of understanding and predicting the physical world's behavior across multiple modalities.
The release of Flux 3 intensifies competition in the rapidly evolving AI video generation space. Market leaders like Seedance have dominated this sector, but multimodal approaches that integrate audio and other data types represent a new frontier.
Flux 3's capability to generate audio natively rather than relying on external tools could prove valuable for content creators, video producers, and enterprises building multimedia applications. The timing of the release positions it within a broader trend of AI models expanding their scope to handle multiple input and output types simultaneously.
Independent benchmarks and user testing will be crucial in determining whether Flux 3's internal performance metrics translate to practical advantages in real-world applications. The availability and pricing structure for Flux 3 remain to be announced.
Runway launched its AI Model Router through Runway Dev on Thursday, positioning itself as an infrastructure layer for generative media rather than just another model company. The router provides API access to third-party image, video, and audio models.
The United States maintains its competitive edge in artificial intelligence through strategic vertical integration by major tech companies, despite ongoing supply chain vulnerabilities and geopolitical tensions with China.
Tech entrepreneurs are urging the Trump administration not to restrict access to Chinese open-weight AI models, arguing that blocking them could harm U.S. innovation and competitiveness.
Poolside released Laguna S 2.1, a compact open-weight coding model that achieves better benchmark performance than significantly larger competitors. The company trained the model to self-correct, revise failed approaches, and persist through extended agentic sessions.