Wednesday night, I was stuck on a client’s product video
At 10:30 p.m., my friend Chen Yu put her coffee cup on the table and asked, “Can you turn one product image into a 15-second video with movement and the sound of a collision?” I’ve been stuck there too: in the past, I had to write a script, find footage, record voice-over, and wrestle with editing software before delivering the next day. This tool isn’t for everyone; not trying it now is fine too. A solid image can still get the job done.
What is it? The two founders made it clear in the demo
FLUX 3, from Black Forest Labs, doesn’t just make images prettier. It understands motion, camera changes, and sound within a scene together. Co-founder Andreas Blattmann used “throwing a ball” as an example: a hand moves before the throw, but we can’t fully describe its trajectory; the model can learn from relationships across frames. Co-founder Dustin Podell left his job to focus on AI creation, and the two have now built these capabilities into the product.
For me, the most practical use is creating a dynamic draft close to a finished video before deciding whether to polish it. Zhang Nan, who runs a personal-IP business from a studio in Hangzhou, storyboarded a jump from an old sofa to a new recliner. That once meant finding an actor, a camera crew, and sound effects. Now, a model can generate a version first so he can test whether the motion feels natural and the sound works. It’s more like a visual assistant that produces concept films than a one-click ad machine.
Recreating it today: modest budget, but you still need visual judgment
Replication cost: Black Forest Labs provides free trial credits, so there’s no need to buy a subscription first. I’d reserve two hours for the first test: 30 minutes learning prompts and 90 minutes generating and reviewing results. The technical barrier is low in terms of coding, but I still need to describe people, motion, camera work, and sound. A normal browser is enough. First step: open the FLUX page from Black Forest Labs, enter the FLUX 3 Video experience, upload a product image, add a motion description, and generate.
I got this wrong too: my first prompt said only “premium and dynamic,” and the result looked like a template ad. Changing it to “the camera moves from left to right toward the cup, the liquid inside sways gently, ending with one crisp sound of glasses touching” produced something useful. I’d keep four versions at a time, compare only whether the motion is coherent and the sound fits the picture, and avoid chasing a longer film immediately.
If I’m busy, this is how I’d choose
Just starting: If I’m still building a personal IP, I’d use the free credits for a five-second animated avatar or product opening and wouldn’t pay to chase new features.
With one or two clients: If I’m repeatedly pitching visual concepts to clients, I’d save three prompt templates: product showcase, person using the product, and before-and-after comparison. That would reduce the need for last-minute communication.
Scaling up: If my team already has scripting and editing workflows, I’d use FLUX 3 for early concepts and filling material gaps while keeping human review. Client budgets, rights, and brand consistency still need someone responsible for the final call.
For me, the safest first step isn’t “use this immediately,” but testing it with a real order: tonight, I’d take one existing product image, generate a 15-second draft, and ask whether the client wants to develop that direction.