Black Forest Labs has moved FLUX 3 Video into general availability, putting its video generation model into the BFL API and making it available through select partners. The release gives developers and creative teams access to a model built for short video clips with sound, multiple input modes, and scene-level control.
What FLUX 3 Video Can Generate
FLUX 3 Video is designed to create HD and Full HD clips up to 20 seconds long. Its output can include native audio, covering dialogue, sound effects, and ambient noise.
That matters because video generation is not only about moving images. For many use cases, sound is part of the finished asset, and FLUX 3 Video includes it as part of the generation process rather than treating it as a separate add-on.
The model supports several creation workflows:
- Text-to-video
- Image-to-video
- Keyframes
- Video continuation
- Multiple scenes and camera angles within a single clip
These options make the model more flexible than a simple prompt-to-clip system. A user can begin from text, build from an image, guide a sequence with keyframes, or extend an existing video. The source also says the model can handle multiple scenes and camera angles within one clip, which points to more structured short-form generation.
Audio, Dialogue, and Scene Details
Black Forest Labs says FLUX 3 Video can generate lip-synced dialogue in more than 14 languages. The company also says the model can render typography directly in scenes, follow complex prompts, and draw on world knowledge for uses such as documentaries.
Those claims place FLUX 3 Video in a category aimed at more than abstract visuals or simple motion tests. Typography inside scenes can be important when a clip needs signs, labels, or other visible writing. Lip-synced dialogue expands the range of narrative and explanatory formats the model can support.
The mention of world knowledge is also relevant for documentary-style uses. Based on the source, BFL is positioning the model as able to follow prompts that require context, not just visual description. The source does not provide examples of those documentary uses, so the practical quality of that capability will depend on real-world testing by users.
BFL Says Its Model Leads In Elo Tests
According to BFL's own tests, FLUX 3 Video ranks at the top of the Elo rankings for both text-to-video and image-to-video. The company reports a score of 1,135 for text-to-video and 1,051 for image-to-video.
BFL says those results place FLUX 3 ahead of Gemini Omni Flash, Minimax H3, and Seedance 2.0. Because the source identifies these as BFL's own tests, readers should treat the ranking as a company-provided benchmark rather than an independent evaluation.
Still, the benchmark claims are central to how BFL is presenting the release. The company is not only making FLUX 3 Video more widely available; it is also arguing that the model is competitive at the high end of current AI video generation.
How FLUX 3 Video Pricing Works
Pricing is based on each second of video output. Audio is included, according to the source.
Draft mode is limited to HD. In draft mode, text-to-video and image-to-video cost $0.06 per second, while video-to-video costs $0.12 per second.
At full quality, HD costs $0.17 per second for text-to-video or image-to-video. HD video-to-video costs $0.41 per second.
Full HD costs more. For Full HD, text-to-video or image-to-video costs $0.29 per second, while video-to-video costs $0.53 per second.
The per-second structure gives users a direct way to estimate cost before generating clips. Since FLUX 3 Video can generate clips up to 20 seconds long, the final price depends on duration, resolution, mode, and whether the user selects draft mode or full quality.
Why General Availability Matters
General availability through the BFL API and select partners makes FLUX 3 Video easier to adopt in products and workflows. Instead of being limited to a narrow test group, the model is now available through channels that developers and partners can build around.
The release combines several features that matter for AI video work: HD and Full HD output, native audio, lip-synced dialogue, image-to-video support, video continuation, and multiple scenes and camera angles in a single clip. For teams evaluating AI video generation, those are the practical details that will shape whether FLUX 3 Video fits their needs.
The most important open question is performance outside BFL's own testing. The company reports strong Elo rankings and positions FLUX 3 Video ahead of Seedance 2.0, Gemini Omni Flash, and Minimax H3, but broader use through the BFL API and partners will show how the model handles everyday prompts, production constraints, and cost-sensitive workflows.