Black Forest Labs Unleashes Flux 3
Black Forest Labs has released Flux 3, a highly anticipated video generation model first teased at Flux 1's August 2024 launch. This new iteration significantly expands creative possibilities, doubling the common video generation limit to 20 seconds. The extended duration directly addresses a key user demand, moving beyond the restrictive 10-second clips typical of many earlier AI video tools.
Flux 3 functions as a robust multimodal model, accepting diverse inputs to drive complex creative workflows. Users can integrate up to 10 references, combining still images, audio tracks, or existing video segments. This powerful capability allows for intricate editing, sophisticated remixing, and precise control over generated content, streamlining advanced production tasks.
Currently, Flux 3 generates video at 720p resolution, with Black Forest Labs planning a 1080p rollout in the immediate future. The model offers comprehensive support for all standard aspect ratios, ranging from vertical 9:16 to cinematic 21:9, ensuring adaptability across platforms. Crucially, Flux 3 also includes integrated audio generation, producing synchronized soundscapes alongside its visual output.
Performance: Coherence Over Kinetics
Flux 3 operates as a reasoning video model, demonstrating advanced prompt coherence. The model successfully executes complex narrative requests, generating outputs that maintain consistent plot elements and character actions over its 20-second generation limit. This capability allows for more intricate storytelling directly from text prompts.
Its performance contrasts with models optimized for rapid visual dynamism. Flux 3 does not prioritize the "hyper pop kinetic energy" or fast action sequences characterized by quick camera movements and trick shots, which are common in some competitor outputs. Instead, it excels in generating dramatic scenes, particularly those featuring extensive, rapid-fire dialogue, an attribute noted as surpassing the pace of "Gilmore Girls."
Black Forest Labs positions Flux 3 as an augmentation tool for professional workflows, not a complete replacement for established solutions such as C-dance. The company has confirmed plans for future release of the model's open weights, facilitating broader integration and customization. Generation costs are projected to be lower than current market alternatives, enhancing accessibility for creators.
AI's Landmark Moment on the Big Screen
Zach London, known as Gossip Goblin, marked a pivotal industry milestone with the theatrical release of his AI-generated feature film. This event represents the most significant distribution achievement to date for AI-driven cinema, transitioning the technology from experimental online showcases to mainstream exhibition. It establishes London as a significant force in the evolving landscape of digital filmmaking.
London's success underscores a growing trend where creator-led initiatives leverage generative AI tools for full-length narrative productions. This development establishes a viable path for independent filmmakers to achieve traditional distribution, demonstrating that AI-generated content can meet the quality and storytelling demands of the commercial film industry. This challenges long-held assumptions about the limitations of machine-generated art.
Theatrical distribution of an AI-generated film validates the technology's capacity for complex narrative construction. For insight into the narrative quality achievable, London’s 30-minute short film, 'Pomegranate,' offers a compelling example. This short demonstrates sophisticated storytelling and character development, indicating the potential for robust cinematic experiences using current AI models; further technical specifications on advanced video generation models, such as Flux 3, are available at FLUX 3 - Real World Models: Towards Multimodal Flow Models as the Backbone of Visual Intelligence. | Black Forest Labs.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
Inside the Gossip Goblin Workflow
Zach London, known as Gossip Goblin, implemented a highly controlled two-step process for his AI-generated feature film, which recently premiered in theaters. The foundational stage involved initial image generation, executed meticulously within Midjourney. This critical first step leveraged specific personalization and S-ref codes, establishing the precise visual groundwork for each scene.
Following image creation, London made a pivotal workflow decision: the exclusive application of first-frame image-to-video conversion. This strategy entailed feeding only a single, pre-designed image into the video generation model for each sequence. Notably, this approach deliberately bypassed multi-referencing entirely, a common practice where multiple images or video clips guide the AI’s output.
London's rationale for this stringent method centered on maintaining absolute creative control. By limiting the AI's reference to only the initial still frame, he exerted precise command over the subsequent animated sequence. This allowed for highly specific direction of camera movement, framing, and shot composition, ensuring the final visual narrative adhered strictly to his artistic vision. This departure from more automated multi-reference techniques underscored a commitment to directorial oversight in AI filmmaking.
Frequently Asked Questions
What is Flux 3 from Black Forest Labs?
Flux 3 is a new multimodal AI video generation model capable of creating clips up to 20 seconds long. It accepts image, audio, or video references and generates its own audio.
What is significant about the Gossip Goblin film?
The film by Zach London (Gossip Goblin) is significant because it's a feature-length, AI-generated movie receiving a theatrical release, marking a major milestone for AI filmmaking.
What is unique about Gossip Goblin's AI film workflow?
His workflow starts with image generation in Midjourney, then uses first-frame image-to-video generation. This avoids multi-referencing to maintain precise camera control.
Is Flux 3 expected to replace models like C-dance?
Initially, Flux 3 is positioned as a tool that can augment existing workflows rather than replace them. It excels in prompt coherence for dramatic scenes but may be less suited for hyper-kinetic action.