Black Forest Labs has taken its FLUX 3 Video model out of testing and made it available to anyone using its API, along with a handful of launch partners. The pitch is simple: type a description, or upload a photo, and get back a video clip up to 20 seconds long, with dialogue, sound effects, and background noise generated automatically inside the same clip.
The model can also render readable text inside a scene, switch between camera angles within a single generation, and produce lip synced dialogue in more than 14 languages. On the company's own internal scoring, FLUX 3 ranks ahead of competing models including ByteDance's Seedance 2.0 and Google's Gemini Omni Flash. That ranking comes from Black Forest Labs itself, so treat it as a marketing claim rather than an independent result until outside testers weigh in.
The pricing is worth pausing on. A 20 second clip at full resolution costs a little under 6 dollars to generate from text, and around 11 dollars if you are reworking an existing video. A comparable professional video ad, with a crew, a studio, and post production, has traditionally cost thousands of dollars and taken days. That gap is the entire reason this category of AI tool exists, and it is why ad agencies and small businesses are paying close attention.
Black Forest Labs is not a random startup chasing a trend. Its founders are the researchers who built Stable Diffusion, the tool that first made AI image generation a household idea back in 2022, before they left Stability AI amid its financial troubles to start their own company. They have since raised a large funding round at a valuation in the billions, backed by investors including Andreessen Horowitz and Nvidia, and signed a multi year contract worth more than 100 million dollars to supply Meta with AI image and video technology. This is a well funded, well connected competitor, not a hobby project.
The video generation market is crowded and moving fast. OpenAI actually shut down its Sora video app earlier this year after facing sustained pressure over deepfakes and videos made of real people without consent. That shutdown is a signal worth remembering: the technical race to make better video is only half the problem, the other half is proving the tool will not be used to fake someone's face or voice without permission.
For businesses, the practical upside is real. Industry surveys already show most ad buyers are using or planning to use AI generated video in their campaigns. But there is a catch: most consumers can now spot AI generated video on sight, and seeing it can lower their trust in a brand. The tools are getting cheap and capable fast. The judgment about when to use them, and when to be upfront that you did, is now the harder and more valuable skill.