The most quietly disruptive thing about FLUX 3 is not that it makes great images, or even that it generates video. It is that you no longer have to choose which tool to open first.
Black Forest Labs, the German AI startup that spun out of Stability AI's founding team, just launched FLUX 3, a single model capable of producing images, short-form video, and synchronized audio from one unified prompt. That is not how the generative AI landscape has worked until now. The current standard involves stitching together separate models — one for visuals, another for motion, a third for sound — and hoping the outputs feel like they belong to the same universe. FLUX 3 collapses that workflow into a single request.
The video output runs up to 20 seconds, which may sound modest until you consider how much of commercial content creation lives in that window. Social ads, product teasers, explainer clips — a huge chunk of what brands actually need falls well within 20 seconds. Pair that with generated audio and you have something that starts to look less like a novelty and more like a production pipeline.
Black Forest Labs is not new to punching above its weight class. Its earlier FLUX.1 models outperformed much better-funded competitors on image quality benchmarks, building the company a loyal following among creative professionals who had grown frustrated with the visual inconsistencies in other tools. FLUX 3 is a significant escalation of that ambition.
The catch, for now, is access. The launch is rolling out in limited release, meaning most people will be sitting on a waitlist while a smaller group stress-tests the model in real conditions. That is a sensible move for a capability this broad — multimodal generation at this level of integration is genuinely new territory, and the failure modes are harder to predict than with single-output models.
What makes the timing interesting is where the broader market is heading. OpenAI, Google, and Meta have all been building toward multimodal generation, but each has largely kept modalities siloed in practice. A smaller, more focused lab beating them to a genuinely unified output is the kind of story the AI industry loves to tell about itself — and occasionally it actually turns out to be true.
For creators and marketers, the more important question is not whether FLUX 3 is technically impressive. It clearly is. The question is whether the quality holds across all three output types simultaneously, or whether the model makes tradeoffs that only become obvious once you start pushing it on real briefs.
That is exactly what the limited release period is designed to find out. If the results hold up under scrutiny, Black Forest Labs will have built something that genuinely changes how creative teams structure their tooling. If not, it joins a long list of demos that looked better in the announcement than in production.
Either way, the bar just moved.
SPACE
SpaceX plans tower catch for Starship after successful 13th flight
Starship's 13th test flight did something none of its predecessors managed: it landed in the Indian Ocean and floated there, intact, like it was waiting for a ride home.
That detail matters more than it might seem. Every previous Starship splashdown ended in flames, which SpaceX had come to treat as an acceptable outcome rather than a problem to solve. This time, the vehicle tipped gently onto its side in a remote stretch of ocean west of Australia and just sat there, heat shield tiles and all, looking more or less like it had when it launched from South Texas roughly an hour earlier. Engineers immediately dispatched drones to get a close look at those tiles, and what they saw was encouraging enough that Elon Musk posted the next mission plan to X before the night was over.
The heat shield is the program's most stubborn engineering problem. Starship's stainless steel airframe is protected by more than 18,000 ceramic tiles that have to survive temperatures approaching 2,600 degrees Fahrenheit during reentry. For SpaceX's long-term vision — rapid reuse, eventually multiple flights per day — those tiles cannot require significant refurbishment between missions. Friday's flight was conducted at higher dynamic pressure than previous attempts, putting more stress on the vehicle during ascent, and the tiles came through. That is real data, not a simulation.
The intact vehicle also opens an unexpected door: SpaceX is now considering towing Starship to shore, likely somewhere in Australia, for the kind of hands-on inspection that drone footage cannot provide. Starship is designed to be reusable, but salt water exposure complicates that, so any recovery operation would need to move quickly.
As for what comes next, Musk made it plain. Unless the post-flight data review turns up something alarming, the 14th Starship test will attempt a tower catch of the ship itself. SpaceX has already pulled this off with the Super Heavy booster, using the mechanical arms on the launch tower — nicknamed Mechazilla — to grab the booster as it descended. Catching the ship is harder. It comes back faster, on a longer trajectory, and the margin for error is smaller.
The plan for flight 14 would also involve sending Starship on a longer-range path, potentially reaching low-Earth orbit, before returning to the launch site. That is a meaningful step toward the operational profile NASA and commercial customers are actually counting on.
It is worth keeping the broader stakes in view here. Starship is the vehicle NASA has contracted to land astronauts on the Moon under the Artemis program. It is also SpaceX's intended platform for Mars missions and next-generation Starlink deployment. Every successful test flight compresses the timeline on all of those commitments.
Flight 13 did not solve everything. But a floating, mostly intact Starship in the Indian Ocean is about the best possible argument that the program is finally moving in the right direction.
Source: Ars Technica
That detail matters more than it might seem. Every previous Starship splashdown ended in flames, which SpaceX had come to treat as an acceptable outcome rather than a problem to solve. This time, the vehicle tipped gently onto its side in a remote stretch of ocean west of Australia and just sat there, heat shield tiles and all, looking more or less like it had when it launched from South Texas roughly an hour earlier. Engineers immediately dispatched drones to get a close look at those tiles, and what they saw was encouraging enough that Elon Musk posted the next mission plan to X before the night was over.
The heat shield is the program's most stubborn engineering problem. Starship's stainless steel airframe is protected by more than 18,000 ceramic tiles that have to survive temperatures approaching 2,600 degrees Fahrenheit during reentry. For SpaceX's long-term vision — rapid reuse, eventually multiple flights per day — those tiles cannot require significant refurbishment between missions. Friday's flight was conducted at higher dynamic pressure than previous attempts, putting more stress on the vehicle during ascent, and the tiles came through. That is real data, not a simulation.
The intact vehicle also opens an unexpected door: SpaceX is now considering towing Starship to shore, likely somewhere in Australia, for the kind of hands-on inspection that drone footage cannot provide. Starship is designed to be reusable, but salt water exposure complicates that, so any recovery operation would need to move quickly.
As for what comes next, Musk made it plain. Unless the post-flight data review turns up something alarming, the 14th Starship test will attempt a tower catch of the ship itself. SpaceX has already pulled this off with the Super Heavy booster, using the mechanical arms on the launch tower — nicknamed Mechazilla — to grab the booster as it descended. Catching the ship is harder. It comes back faster, on a longer trajectory, and the margin for error is smaller.
The plan for flight 14 would also involve sending Starship on a longer-range path, potentially reaching low-Earth orbit, before returning to the launch site. That is a meaningful step toward the operational profile NASA and commercial customers are actually counting on.
It is worth keeping the broader stakes in view here. Starship is the vehicle NASA has contracted to land astronauts on the Moon under the Artemis program. It is also SpaceX's intended platform for Mars missions and next-generation Starlink deployment. Every successful test flight compresses the timeline on all of those commitments.
Flight 13 did not solve everything. But a floating, mostly intact Starship in the Indian Ocean is about the best possible argument that the program is finally moving in the right direction.
Enjoyed this?
Get stories like this delivered every Tuesday — free.