Previously, I looked at using the Linux video loopback system from the command line. The basic trick was simple enough: capture video from a real camera, process it with something like ffmpeg, and write the result to a fake camera device via the v4l2loopback device. Then a browser, or any camera-enabled software, sees the fake camera as if it were real. This allows you to manipulate video before sending it to the rest of the world.

That works, and for those of us who like command lines, it’s easy enough to execute. But not everyone loves the command line. In the comments, there was another obvious answer: use OBS Studio.

While OBS is excellent, it is also a bit like using a laser to chop a carrot. If you already use OBS, fine. If you only want to crop a webcam, add an effect, mirror an image, or feed a virtual camera, it can feel like a lot. If you must have a GUI, you can try Webcamoid, which sits somewhere between a simple webcam viewer and a full video production system.

Webcamoid gives you a GUI for selecting a camera, applying effects, and sending the result to a virtual camera. Conceptually, it is much closer to the command-line loopback setup from the previous post than to OBS. You are still building a pipeline from input camera to output camera, but now you can do much of it with buttons and menus instead of shell commands.

That’s in theory, of course. Implementing Webcamoid turned out to be quite the exercise. Granted, this probably varies depending on where you install software. If your distro has a clean working copy of Webcamoid and its dependencies, good for you. For everyone else, keep reading.

The Moving Parts

There are two pieces to understand. Webcamoid is the GUI application. It captures video, previews it, applies effects, and can write to a virtual camera. You also need a driver to produce the fake or virtual camera. AkVCam is one of the virtual camera drivers that Webcamoid can use on Linux. It can also use v4l2loopback, as we discussed last time. Both approaches create fake /dev/videoX devices, but their configuration models are different.

With v4l2loopback, the typical setup is command-line oriented:

sudo modprobe v4l2loopback video_nr=10 card_label="Virtual Camera" exclusive_caps=1

Then some program writes frames to /dev/video10. For example, ffmpeg can read from a real webcam and write to the virtual one. Of course, if you want a permanent virtual camera, you can make an entry in /etc/modprobe.d.

AkVCam is more structured. Instead of simply creating a generic loopback device, it uses a configuration file that defines one or more virtual cameras, their input/output relationship, and the formats they support. That sounds like more work, and in a way it is. But it also gives you tighter control over what formats the virtual camera advertises. This sometimes matters more than you might expect.

Install, Remove, Install…

The hard part of Linux webcam work is often not getting video — it’s getting every piece of the chain to agree on width, height, frame rate, and pixel format, along with matching each other’s API expectations.

I tried three different ways to install Webcamoid. First, I used the normal OpenSUSE Tumbleweed repos to install the program. It couldn’t find any cameras. My next stop was a Flatpak version. That worked well, but it is deliberately crippled and won’t even try to drive a virtual camera, directing you to install the regular version instead. Then I tried an AppImage. This seemed to work OK, but the virtual camera would never display anything but a black screen.

Note that the version on AppImageHub is old, and the source project requires payment for prebuilt binaries. I didn’t try either of those options.

I tried a lot of things to make it work. My final answer was to use the AppImage, but I had to build my own version of AkvCam from GitHub.

Even then, at first, the video output was highly pixelated. The culprit was AkVCam using the 640×480 RGB format. Upscaling created a blocky mess.

You can see what a device is doing with:

v4l2-ctl -d /dev/video3 --all

In my case, the virtual output reported:

Width/Height : 640/480
Pixel Format : 'RGB3'
scaling_mode : Fast
aspect_ratio_mode : Ignore

That explains it. “Fast” scaling usually means “not pretty,” and 640×480 is not a great starting point for modern video calls.

The fix was to simplify the AkVCam configuration. Instead of giving the virtual camera a long list of supported formats, I configured it with essentially one useful format. For example, if the pipeline is meant to be 1280×720 at 30 fps, make that the format. Do not give every program in the chain an opportunity to negotiate itself down to postage-stamp resolution.

A minimal AkVCam setup generally defines a capture device, an output device, a connection between them, and formats. In the generated config I was using, the output camera had several formats:

cameras\2\type=output
cameras\2\videonr=3
cameras\2\formats=19, 20, 21, 22, 23, 24

The problem was that format 19 was 640×480. The high-resolution formats were present, but not preferred. Reordering the list might help, but for reliability, using only the desired format is even better.

Webcamoid As A GUI Pipeline

Once the virtual camera driver is sane, Webcamoid works well. You select the real camera as the source, apply whatever effects you want, and select the AkVCam virtual camera as the output. The receiving application then sees the virtual camera.

Compared to the command-line approach, Webcamoid makes experimentation easier. Want to flip the image, adjust color, crop, blur, or test silly filters? That is all much easier in a GUI than in an ffmpeg or gstreamer filter chain, although I still don’t mind the command line. There’s not much difference between using Webcamoid as a friendly front end or manipulating the image as we did last time.

However, I would not oversell it. Webcamoid is not OBS. It is not really a scene compositor. If you want multiple live sources — say, a screen capture as the main image and your webcam as picture-in-picture — OBS is the obvious GUI tool. With the command line, you can easily do things like this by calling ffmpeg directly.

That’s a Wrap!

If you want a full production environment, OBS is still the right answer. It handles scenes, multiple sources, transitions, screen capture, overlays, and virtual camera output in a single application. But if you liked the command-line loopback idea and wished it had a friendlier face, Webcamoid plus AkVCam is worth a look. It gives you a GUI for the common case: one camera in, effects applied, one virtual camera out. You just need to fight through the installation and configuration with your specific setup. Hopefully, yours will be easier than mine.

As usual, Linux rewards knowing what is happening under the hood. Webcamoid can make the workflow friendlier, but v4l2-ctl is still your friend, and, at least for some people, you’ll need some Linux Fu to get it all working in harmony.

• A new Entertainment Software Association study found that most Americans think video games provide the most entertainment value for their money compared to other forms of entertainment
• 75% of parents are actively play video games each week, while 81% enjoy playing with their children
• The majority of players across all age groups are also spending money on in-game content

A new Entertainment Software Association (ESA) study has found that the majority of gamers in the United States prefer to spend their money on video games because they think they provide more entertainment value than other forms of media.

As detailed in its annual Essential Facts About the U.S. Video Game Industry report conducted in partnership with YouGov, video games are more popular than ever among most age groups.

from the good-deals-on-cool-stuff dept

LabsDigest is built for those who learn best by doing. Whether you’re preparing for a CompTIA certification or diving into Python development, our platform offers interactive labs that simulate real-world tasks—no passive watching or reading, just real experience. Work through performance-based exercises for CompTIA A+, Network+, Security+, and more, or sharpen your coding skills with hands-on Python projects. Learn by solving problems, fixing bugs, and applying your knowledge in practical scenarios that prepare you for the real world. It’s on sale for $30.

Note: The Techdirt Deals Store is powered and curated by StackSocial. A portion of all sales from Techdirt Deals helps support Techdirt. The products featured do not reflect endorsements by our editorial team.

Filed Under: daily deal

The Swedish app-builder, processing a million new projects a week, is making Google Cloud a primary partner, with Gemini models and a security layer aimed at corporate buyers.

The pitch behind Lovable has always been that anyone can build software by chatting with an AI. The harder pitch, the one that turns a viral tool into a durable business, is that a large company can trust what gets built.

On 3 June, at Google Cloud’s Nordics summit in Stockholm, Lovable set out to make that second case, announcing an expanded multi-year collaboration with Google Cloud aimed squarely at enterprise buyers.

The deal makes Google Cloud one of Lovable’s primary technology partners, anchoring its platform on Google’s AI infrastructure and Gemini models. Lovable says its users are now processing more than one million new projects every week, a volume that has outgrown the scrappy infrastructure a consumer tool can run on and needs the secure, enterprise-grade backing a hyperscaler provides.

Lovable is one of the more remarkable growth stories in recent software. Founded in Sweden and built around what the industry has taken to calling “vibe coding,” turning natural-language prompts into full-stack applications, it raised a $200M Series A in mid-2025 at a $1.8bn valuation and was reported to be valued at around $6.6bn by the end of the year. The company says builders created more than 25 million projects in its first year, and that Lovable-built applications now draw 600 million visits a month.

The collaboration is built on three pillars, and they read as a checklist of what enterprises demand before they let an AI tool near production. The first is a verified agent: Lovable has launched its Lovable Agent in Google Cloud’s Gemini Enterprise Agent Gallery, a vetted catalogue of third-party agents that corporate customers can adopt with some assurance about what they are running.

The second is security, reinforced by a new integration with Wiz, the cloud-security company Google is acquiring, to identify and remediate vulnerabilities in AI-generated code in real time, alongside continuous scanning, dependency checks, permissioning and audit trails.

The third pillar is the least glamorous and arguably the most telling: simplified procurement and billing through Google Cloud Marketplace and Gemini Enterprise.

Enterprises buy software through approved channels with predictable invoicing, and being available where a corporate buyer already has a billing relationship removes a quiet but real barrier to adoption. The pillar exists because procurement, not capability, is often what stalls enterprise deals.

The security emphasis is the substantive part of the announcement, because it speaks to the central anxiety about AI-generated code. Tools that let non-engineers ship applications also let them ship vulnerabilities they cannot see, and for a regulated enterprise that risk is disqualifying.

Wrapping Lovable’s output in continuous scanning and remediation is the company conceding that “anyone can build” needs “and it will be checked” attached before a serious buyer signs on.

There is a competitive subtext worth noting. Lovable sits in a crowded vibe-coding field alongside Cursor, Replit and Bolt, and the AI model providers themselves are building rival app-creation tools.

Tying tightly to Google Cloud, and to Gemini, gives Lovable a hyperscaler’s distribution and infrastructure at a moment when its rivals are racing for the same enterprise budgets. It also slots into Google’s broader campaign to win the “agentic enterprise,” the same push behind its $750M partner fund for agentic AI.

As a Google Cloud announcement, the framing is naturally Google’s, and the deeper commercial terms, what each side pays and commits, are not disclosed. What the partnership establishes is direction.

Lovable has decided its next phase runs through the enterprise, and that getting there means less talk of how easy building is and more proof that what gets built is secure, governed and accountable. The million projects a week are the easy part. Convincing a Fortune 500 compliance team is the part this deal is for.

Most larger ride-around landscaping machinery has a similar transmission, a transaxle containing a gearbox, or in some cases, a continuously variable drive. [Made In Garage] has a Toro lawn tractor with just such a setup, and when the transaxle failed he replaced it with a hydraulic drive.

The video below is a classic bit of workshop porn, as he fabricates both the hubs and the rear frame to fit a pair of hydraulic motors. The throttle pedal is a hydraulic valve with the lever swapped for a pedal, and the hydraulic reservoir, in a nice touch, is an old fire extinguisher.

We’re not so sure about the pipework in such an exposed position under the machine as we think it would inevitably be damaged, but you can’t argue with the results. Having used a rough service mower with a hydraulic drive in the past, we appreciate always being exactly at the right ratio for the engine.

We think perhaps he should complement it with a loader.

Thanks [Keith Olson] for the tip!

from the there-is-no-magic-wand dept

Before the current wave of laws banning mobile phones in schools, we had published a piece from some researchers who had looked at how similar bans had worked in Australia, with the conclusion that… they didn’t. At best, the research showed the evidence on school phone bans to be “weak and inconclusive.” Those authors suggested that rather than doing outright bans, politicians should leave the issue to the schools themselves to determine what’s best.

So it should come as little surprise that two years later, after many similar bans have gone into effect in the US that… the studies are showing up as (you guessed it) weak and inconclusive. The new study from the National Bureau of Economic Research (NBER) has some people shaking their heads because it can find no evidence of better student performance in schools.

Schools that adopted strict bans — requiring students to keep their devices in locked pouches throughout the school day — saw a meaningful decline in student cellphone use. But test scores have not increased in those places on average. And at first, banning phones led to higher suspension rates.

That’s not to say there should be a free for all in schools. But, once again, it would be nice if politicians, the media, and other commentators could finally (for once) recognize that blanket bans of technology are almost never the answer. The relationship between students and technology is complex and nuanced and doesn’t have a single effect in a single direction. Instead, it’s highly context and individual dependent.

A reasonable, nuanced approach is (1) better equipping teachers with tools to be flexible, (2) better educating students on the tradeoffs of technology use, and (3) improving the overall education environment with an actual recognition that context matters.

Obviously, if kids are just sitting in class all day staring at their phones instead of paying attention to the teacher, that’s a problem. But there are ways to deal with that specific scenario that don’t require a full ban. For some schools a full ban could absolutely make sense, and for others it doesn’t.

We’re going through this in our own local school district where, starting a few years ago, the high school (after studying and then testing different solutions) put boards in every classroom with pockets, where students were asked to deposit their phone at the beginning of every class. This created some challenges, such as when some teachers used the phones in pockets as an attendance-taking short-cut and some students (including my own kid) did not have a phone with them at all in school (by their own choice at the time). But it also meant kids could have phones on them between classes and at lunch.

It’s not a perfect solution, but that’s an important point: nothing is a perfect solution, and pretending otherwise is a problem.

But then California passed a new law, which required schools to come up with plans to ban phones. While the law was not nearly as strict as many other school phone ban laws and does actually give schools more freedom in creating a policy for their own community, it already has resulted in a bunch of wasted time where our school district feels they need to go back to the drawing board and come up with a new phone ban plan, even though the old one appeared to be working decently well.

At the last minute, California has even scaled back its original law, giving schools even more freedom — but also more confusion.

But this whole episode reeks of the usual political and media reflex of “we must do something, this is something, therefore we will do it.”

Letting communities and schools decide how best to handle in school distractions seems like a much more appropriate approach. And part of that is teaching everyone that there is no magic bullet solution to problems. Kids are in school to learn, and part of that learning should be “hey, you shouldn’t be staring at your phone all day, but also it shouldn’t take a law to get you to put down the phone.”

The new study just confirms what the earlier research already showed. Blanket bans make for good press releases. They don’t make for better students.

The hard, unglamorous work of actually improving education — better-equipped teachers, more engaged classrooms, students who’ve been taught to think critically about their own technology use — doesn’t fit on a campaign mailer. So we keep getting laws instead of solutions.

Filed Under: california, education, phones, school phone bans, schools, teens

Eighteen months ago Suno was the AI company the music industry wanted to destroy. Every major record label had sued it, accusing it of training its models on copyrighted songs without permission. Now the labels are its partners, and investors have repriced the company accordingly. Suno has raised new capital at a $5.4bn valuation, more than double the $2.45bn it was worth just six months ago.

Bond Capital led the round, a Series D that had been reported to be closing for several weeks. The step-up is steep: a little over 2x in roughly half a year, the kind of re-rating that usually reflects either explosive growth or a fundamental change in a company’s risk profile. In Suno’s case it reflects both, and the second may matter more than the first.

The growth is real enough. Suno says more than 100 million people have now used the service, with around 2 million paid subscribers, and it reported roughly $150M in revenue in 2025.

By the measure investors care about, it is one of the breakout consumer-AI products, turning text prompts into finished songs for a mass audience rather than a niche of producers.

The 💜 of EU tech

The latest rumblings from the EU tech scene, a story from our wise ol’ founder Boris, and some questionable AI art. It’s free, every week, in your inbox. Sign up now!

But the more important shift is legal. When Suno last raised, it did so under an existential cloud: lawsuits from Universal, Warner and Sony, any of which could in principle have ended the business if the courts found its training data infringing. That cloud has substantially lifted.

Warner settled in November 2025 and struck a partnership to build licensed models, and Universal settled in October, its deal pairing a payment with a licensing arrangement for a joint AI platform. Two of the three majors that wanted Suno gone are now commercial partners.

That is the change investors are paying for. A company facing three industry-ending lawsuits is priced for the possibility of zero. A company that has converted two of those plaintiffs into licensors, and is rebuilding its models on authorised catalogue with artist opt-in, is priced as a going concern with a path to legitimacy. The valuation jump is less a bet on more users than a re-rating of the chance that Suno survives to keep them.

The terms of the settlements point at what Suno becomes on the other side. It has said it will launch new, licensed models in 2026 and deprecate the current ones, give artists and songwriters control over whether their names, voices and compositions are used, and require a paid account to download audio.

That is a more constrained, more expensive product than the free-for-all that built its user base, and the strategic question is whether 100 million users trained on the old model accept the new one.

The truce is also incomplete. Sony, the last of the three majors still litigating, has settled with neither Suno nor its rival Udio, and its fair-use cases are expected to produce a pivotal ruling in summer 2026. That decision could shape the copyright ground rules for the entire generative-music field, and a result unfavourable to Suno would complicate the legitimacy narrative this valuation rests on. The risk has shrunk, not vanished.

What the round captures is a company mid-transformation, from insurgent to licensee, from sued to partnered, from free tool to paid platform. The $5.4bn says the market believes the transformation is working. Sony’s lawyers, and a courtroom this summer, still get a say in whether it finishes.