Sponsored by
Hey Creator,
Until now, FLUX has been your go‑to for images. With the latest update, Black Forest Labs is treating FLUX 3 as one backbone for video with synced audio, next‑gen images, and robot action.
Here’s what actually changes for creators — and what’s still early access.
One Model, Four Jobs
If you've used FLUX for image generation, you've been working with one piece of a much bigger model. On July 23, Black Forest Labs showed what the rest of it was built for.
The company launched FLUX 3 — built to generate video with sound, edit images, and even predict robot movement, all from the same underlying network.
What makes this different from FLUX.1 and FLUX.2:
Earlier FLUX models only did images. FLUX 3 is trained jointly across image, video, and audio instead of stitching three separate models behind one interface.
The idea: an image only shows a single instant, video adds motion, and audio is what ties the two together — the sound has to match the impact, the movement has to obey physics. Training all three together, rather than one at a time, is meant to make each of them more accurate on its own.
The model comes in four parts:
FLUX 3 Video — generates clips up to 20 seconds with synchronized audio built in, from text, images, or an existing video
FLUX 3 Image — the next generation of FLUX's image tool, rolling out in the coming weeks
FLUX 3 Action — extends the same backbone to predict robot movement, currently being tested with manufacturing partners
FLUX 3 Dev — a planned open-weight release of the same backbone, no date confirmed yet
How it's actually performing:
Black Forest Labs shared its own early test results: FLUX 3 was preferred over Runway Gen-4.5 in 77% of comparisons, over Luma Ray 3.2 in 93%, and beat Grok Imagine Video in up to 69%. Against Kling and Seedance, the margins were tighter, in the 50–60% range.
Worth knowing before you get excited: these are Black Forest Labs' own numbers, not independently run, and the tests were on 10-second clips at 720p — shorter than the full 20 seconds the model can produce. Treat the percentages as a starting signal, not a settled comparison.
Getting access right now:
FLUX 3 Video and FLUX 3 Action are in gated early access — you can apply at bfl.ai/models/flux-3, but Black Forest Labs approves each request individually. Image access follows in the coming weeks. There's no public API and no pricing yet, so it's too early to plan a production budget around it.
One thing worth flagging directly: copycat sites using names like "flux-3.com" have already shown up and have nothing to do with Black Forest Labs. The only official source is bfl.ai.
Why this is worth watching if you edit video:
The appeal, if it holds up outside the company's own testing, is skipping the separate audio pass entirely — generating a clip where the sound already matches the motion, instead of layering it in afterward.
That's real time back for anyone doing short-form video, reels, or product clips, where matching sound to cuts is often the slowest part of the edit. It's still behind an approval gate right now, so this isn't a workflow to build around yet — just one to keep an eye on.
Data flag: Performance comparisons and technical details here come from Black Forest Labs' own announcement, not independently verified benchmarks.
Cap table management that works for you
Managing your cap table shouldn’t drain your time or derail your budget. Pulley takes the complexity and surprises out of equity management with intuitive workflows, audit-ready compliance, and accurate reporting—all backed by support and pricing you can rely on.
CFOs and finance leaders trust Pulley to stay compliant with key tasks like 409A valuations, ASC 718 reporting, or 83(b) elections without incurring expensive legal fees or getting bogged down with manual work. With Pulley, you get 100% audit-defensible data, powerful tools, and real expertise.
