For most enterprises, a 90-second training video or a product explainer has never been an easy ask. It means a well planned brief, an internal film crew or an outside vendor, a shoot, an edit, and a round of revisions. Change one line of on-screen text due to a legal review and the whole chain runs again. The cost and the long time lines are why so much internal video never gets made.

That equation is what Google is aiming to rewrite with Gemini Omni Flash, the first model in its new “Omni” family, now rolling out to developers and enterprise customers through an API after debuting to consumers at I/O 2026. Google frames the family’s ambition as creating anything “from any input,” starting with video. But the headline interaction isn’t just a sharper text-to-video prompt. It’s the ability to edit a finished clip through conversation.

When the model launched in May, VentureBeat’s enterprise analysis flagged the catch: with no programmatic interface, Omni was a consumer and prosumer tool, not a production one. This API rollout changes that. It puts conversational editing in front of the marketing and learning-and-development teams that make the most videos in an organization.

The pitch: a five-tool pipeline collapses into a single conversation

Until now, many teams have been assembling AI videos the hard way, bolting together an LLM for a script, a text-to-image model, an image-to-video model, a separate lip-sync tool and a voice generator, each with its own contract, billing and data path.

Omni’s enterprise argument is unification: one model that takes text, images and video and returns a finished clip with synced audio.

That simplicity factor is the part decision-makers should weigh first. Collapsing several point tools into one model means fewer vendors and a single place to monitor output and enforce data-handling rules. For an organization that has avoided generative video because stitching the tools together wasn’t worth the overhead, the equation shifts.

With conversational editing each instruction builds on the last, so a marketer can relight a product shot, reframe it, or change the wardrobe without regenerating from scratch and losing the parts that already worked. It is the difference between booking a reshoot and sending a note.

Multimodal references and a physics engine for brand assets

Omni accepts far more than a text prompt. Alongside the words describing what you want, you can feed it multiple reference images, and existing video clips, and it carries those specifics into the result. Hand it a photograph of a particular object, ask the model to place that object into a scene, and it reproduces the real thing’s coloring and rough shape instead of inventing a generic stand-in. While the match might not be pixel-perfect, it is close enough to be recognizable. That reference-driven control is what makes the feature commercially interesting: a product photo, a brand logo, or a specific location can be dropped in as an ingredient rather than described in a prompt and hoped for.

Two of Google’s four highlighted strengths speak directly to enterprise work. The first is a world model, the system’s grasp of how physical scenes behave. Add light rain and puddles to an existing shot and it renders reflections of the people and objects in the wet pavement, the sort of physical consistency that separates real footage from obvious AI video. 

The second is text and logo insertion. Point it at a scene full of signage and you can have it rewrite those signs in another language, or for a brand of your choosing, and even drop in a company’s logo. The results aren’t flawless: in testing, sign tracking in complex scenes weren’t always perfect and some text slipped back to the original language between frames. For training videos that need on-screen labels, or ads that need a logo placed in-scene, it is a capability worth a close look, and a reminder that the output still needs a human review before it ships.

The interactions API and where the limits still bite

Under the hood, this runs on Google’s new interactions API, a stateful interface built for multi-turn tasks rather than open-ended chat. Each turn carries the previous video and its references forward, which is what lets edits accumulate coherently. Developers can chain generations. They can produce a clip, edit the cat into a puma kitten, restyle a video into 8-bit retro and then into a watercolor look, and store each version to branch from later.

The constraints are real and worth budgeting around. Clips currently cap at 10 seconds, per the model’s published model card. To make something longer, you generate chunks and edit them together. Uploaded footage can be edited too, as long as it runs 10 seconds or under and the user holds the rights to it. Google’s own model card is candid that holding consistency across edits and rendering accurate text remain open problems.

Guardrails, watermarking and the line Google won’t cross

For a CISO, the demos matter less than the provenance work shipping alongside the model. Every Omni clip carries Google’s SynthID watermark, Google is extending C2PA Content Credentials across its generative tools, and it has launched an AI Content Detection API that flags AI-generated media, both Google’s and other vendors’.

Google has also drawn a deliberate line. The model won’t take a still photo of a person plus an audio clip and lip-sync them into speech, an explicit move to limit deepfakes. It will, however, take a recording of someone talking and translate it into another language, a useful path for localizing global training content. For regulated enterprises, those constraints and the baked-in provenance are features rather than friction.

The numbers: cheap, 720p-only, and (preliminarily) ranked first

The pricing landed alongside the API, and it is aggressive. Omni Flash costs $0.10 per second of generated 720p video, which puts a ten-second clip at roughly a dollar. That matches Veo 3.1 Fast at the same resolution, runs double Veo 3.1 Lite, and undercuts standard Veo 3.1 by three-quarters.

The table also exposes the catch though. Omni Flash only generates 720p. There is no 1080p or 4K option, while the Veo tiers scale up to 4K. For internal training and most social video, 720p is fine. For premium brand work meant for a large screen, it is a real ceiling, and the reason Veo 3.1 still has a job

Clips run 3 to 10 seconds at 720p native, in landscape (16:9) or portrait (9:16). As reference inputs the model accepts up to seven images and up to three video clips of three seconds or less. It does not take audio as an input yet, though it generates audio alongside the video it produces. Output is standard MP4, and every clip ships with SynthID watermarking and C2PA credentials baked in.

On quality, the early signal is strong. In LMArena’s Text-to-Video Arena, a leaderboard where people vote on head-to-head outputs from competing models, Omni Flash sat at number one with a score of 1527. 

What it means for budgets, and what’s still missing

With real pricing in hand, the iteration story gets concrete. Every conversational edit is a fresh generation you pay for, so an edit-heavy session still adds up, roughly a dollar for each ten-second pass at 720p. What the stateful model changes isn’t the cost of an edit, it’s the number of wasted ones: because context carries across turns, those generations go toward refining a take that mostly works instead of restarting from a blank prompt and hoping the next attempt lands.

Omni isn’t alone in this field. Veo 3.1 remains Google’s production-grade option when you need higher resolution, and rivals from Bytedance, Alibaba and OpenAI are all chasing the same budgets. What Omni adds is the editing capability itself: the ability to treat a video as a living document instead of a one-shot render.

WhatsApp’s long-awaited username feature is now officially rolling out to users. But almost as soon as it was announced, many began asking an obvious question: won’t this make it easier for scammers to message strangers? Now, WhatsApp has stepped in to explain why it believes that won’t happen.

WhatsApp says usernames aren’t as open as Telegram’s

Much of the concern stems from comparisons with Telegram, where anyone can search for a public username and immediately start a conversation. Several users on X argued that hiding phone numbers improves privacy but also removes a layer of accountability that helped identify suspicious contacts.

As the rollout began, WhatsApp responded directly to users on X, explaining that its implementation works very differently. For starters, there won’t be a public directory or username suggestions to help people discover accounts. Instead, someone will need to know your exact username before they can even try to contact you.

The company also revealed another privacy layer called a username key. If users choose to enable it, nobody can message them using their username unless they also know that key, adding an extra hurdle for unwanted messages. WhatsApp says it has built several anti-abuse measures into usernames from day one. The company will rate-limit how many new people an account can contact, block repeated attempts to guess someone’s username key, and use existing systems to detect and remove impersonation or other suspicious activity.

Furthermore, even if someone does message you, WhatsApp says the app will continue to provide useful context, including whether the sender is a new account, already in your contacts, shares a mutual group with you, or is based in another country. Users will still have the same options to block, report, or ignore unwanted conversations.

Privacy comes with new responsibilities

The funny thing is that WhatsApp’s biggest challenge isn’t the technology; it’s changing user habits. On most social platforms, people try to grab a username that matches their real name. While WhatsApp emphasizes there won’t be a public directory to browse, using your real name could still make your handle easier to guess. If privacy is the ultimate goal, choosing a more unique username may be the smarter move.

As usernames gradually roll out to more users, it’ll become clearer how well these protections hold up in the real world. But one thing is already clear: WhatsApp knew the scam concerns were coming, and it has designed usernames to prioritize privacy over discoverability, making them far less open than many users initially feared.

Every experienced machinist knows the value of taking regular measurements. If one works carefully and checks dimensions frequently, it’s possible to make a part much more precise than could be made by relying on the machine’s accuracy alone. In a similar vein, it’s possible to make a measuring device out of comparatively crude parts, as long as their behavior is well understood. Related to both principles is [BubsBuilds]’s displacement sensor, which uses a 3D printed frame but reaches precision better than two micrometers.

Admittedly the printed parts aren’t the source of the sensor’s precision, that comes from an opto-interrupter. This design has a central stylus, one end of which contacts the object under measurement. The other end flattens to a knife-edge blade, which fits between the diodes of the opto-interrupter. As the stylus point is pressed in, the blade blocks off more light from reaching the photodiode, creating an output signal proportional to displacement. To keep the stylus from twisting or moving side-to-side, two flat, circular flexures hold the stylus in the center of a cylindrical housing.

[Bubs] printed several flexure variations to see how well they resisted and permitted various torques and forces, and a symmetrical flexure design proved best for his purposes. Once the sensor was assembled, he tested it against the measurements recorded by a laser confocal displacement sensor. This design was an update from a previous version, and it improved in a few regards: the non-linearity had decreased, and the repeatability was now better than two microns, though the range had been halved. Significantly, though, it’s now much easier to mount, making this an actually practical tool.

If, however, this doesn’t fit your needs, there are many other ways to build a linear displacement sensor, ranging from capacitive to magnetostrictive. On the manual side of things, we’ve also covered a comparison of calipers.

Clicks just released its first hands-on video of a working Communicator prototype this morning. The clip shows pre-production hardware running actual software, with marketing lead Jeff Gadway demonstrating calls, messaging, music playback, and app navigation on the device. Earlier appearances relied on non-functional dummy units. This version moves the project from concept to something people can picture using every day.

The form size is kept purposely small, with a 4-inch OLED screen atop a complete physical keyboard that takes up the majority of the front of the device. The white variant appears to be very clean and purposeful, with pill-shaped keys that are slightly elevated, making them easy to locate by touch. The spacebar at the bottom is lovely and broad, and it even has a fingerprint sensor built in, which fits your thumb’s natural resting position, allowing you to unlock the device without having to move your hand away from the keyboard.

Sale

Samsung Galaxy S26 Ultra, Unlocked Android Smartphone, 256GB, Privacy Display, Galaxy AI, AI Camera…

• PRIVACY DISPLAY: Automatically hide your screen from those beside you. The built-in privacy display can be preset¹ to turn on when receiving…
• TYPE IT IN. TRANSFORM IT FAST: Enhance any shot in seconds on your smartphone by using Photo Assist² with Galaxy AI.³ Add objects, restore details…
• NIGHTS, CAPTURED CLEARLY: From gigs to city lights, record and capture moments after dark with clarity using Nightography so your photos and videos…

You’ll find a variety of essential functionality scattered around the hardware. On top, there’s a 3.5mm headphone jack next to one microphone, with two additional at the bottom and back to help with calls as well as recordings. A barometric pressure sensor is also included to aid with position accuracy and other functions. You can charge it using USB-C or wireless charging, and the 4000mAh battery should be enough for light daily use. It sports a good 50-megapixel main sensor and a 24-megapixel front camera for quick shots or video calls with another device.

The removable rear cover is an excellent design choice that is also extremely user-friendly. Simply remove the panel with your finger through a little notch and some chamfered edges, and voilà! Inside, there is a SIM card slot and a microSD card reader that can accommodate 2TB cards, all of which are easily replaced.

The prototype runs Android 16 with a modified UI based on the Niagara launcher. The home screen is relatively basic, with a ribbon of favorite apps over on one side so you can quickly grab what you need, and app notifications are integrated into the main view rather than obnoxious floating banners, which is good! To respond to a message preview, simply swipe it. Yes, typing on the actual keyboard is also fairly slick, as it searches for apps and content and displays results right away. The home screen also includes several widgets, like as playback controls for your music apps.

Overall, Clicks sees this small handset as a companion rather than a full-fledged substitute for larger flagship phones. Many individuals will maintain their large cellphone for taking images or using demanding apps, but this small unit is ideal for quick messages, short conversations, and focused notes. If you simply want a phone that is less flashy and allows you to work without being distracted, this could be an excellent choice. It has all the necessary connectivity, including 5G, Wi-Fi, Bluetooth, and NFC.

Shipping is scheduled for the fourth quarter of 2026, with a target price of $499. They’re also taking pre-orders on their website. This demonstrates that the team has made significant progress, as the transition from a static display model to something that actually does the real thing is substantial.

Anthropic today released Claude Sonnet 5, a new AI model that the company says delivers near-flagship performance at mid-tier prices — a move designed to give cost-conscious enterprise developers access to powerful agentic capabilities just as the San Francisco-based AI lab barrels toward an initial public offering that will test whether the private market’s staggering AI valuations can survive public scrutiny.

The release, which Anthropic describes as “the most agentic Sonnet model yet,” makes Sonnet 5 the default model for users on Anthropic’s Free and Pro plans, while also making it available to Max, Team, and Enterprise customers. Introductory API pricing is set at $2 per million input tokens and $10 per million output tokens through August 31, after which it rises to $3 and $15 respectively — still well below the $5 input and $25 output pricing of Anthropic’s top-of-the-line Opus 4.8.

The strategic logic is unmistakable: Anthropic is trying to democratize access to capabilities that until very recently only its most expensive models could deliver, while building the kind of broad-based developer adoption that will look attractive in an S-1 filing.

Sonnet 5 benchmarks show the mid-tier model closing in on Anthropic’s flagship Opus

Sonnet 5 posts major gains over its predecessor, Sonnet 4.6, across every evaluation Anthropic disclosed. On SWE-bench Pro, an agentic coding benchmark, Sonnet 5 scores 63.2% compared with Sonnet 4.6’s 58.1% — a jump that brings it within striking distance of Opus 4.8’s 69.2%. On Terminal-Bench 2.1, another coding evaluation, the gap narrows further: 80.4% for Sonnet 5 versus 67.0% for Sonnet 4.6 and 82.7% for Opus 4.8.

In multidisciplinary reasoning, as measured by Humanity’s Last Exam, Sonnet 5 scores 43.2% without tools and 57.4% with tools — the latter figure essentially matching Opus 4.8’s 57.9%. On computer use tasks evaluated through OSWorld-Verified, Sonnet 5 reaches 81.2%, up from 78.5%. And on GDPval-AA v2, a knowledge-work benchmark, it scores 1,618 — surpassing Opus 4.8’s 1,615 and far exceeding Sonnet 4.6’s 1,395.

The pattern across these evaluations tells a consistent story: Sonnet 5 doesn’t merely inch forward from its predecessor. It vaults into a performance tier that overlaps substantially with Anthropic’s flagship model, while costing roughly 60% less per token at standard pricing and even less during the introductory period.

Enterprise partners say Sonnet 5’s agentic AI capabilities finish jobs that previous models abandoned

The emphasis on agentic capabilities — the ability to plan, use tools like browsers and terminals, and execute multi-step workflows autonomously — reflects where the AI industry’s center of gravity has shifted in 2026. Enterprises are no longer simply asking chatbots questions; they are deploying AI systems that can navigate complex software environments, execute multi-step coding tasks, and operate with minimal human supervision.

Early access partners painted a picture of a model that doesn’t just start tasks but finishes them. Sualeh Asif, co-founder of Cursor, the AI-powered code editor that has become a bellwether for developer tool adoption, said that “with Claude Sonnet 5, agents stay on plan, follow our conventions, and ship clean multi-step changes, all at an efficient cost.” Daniel Shepard, a senior engineer at Zapier, described handing the model a two-part automation job — updating Salesforce account tiers and sending a launch announcement — that “used to stall halfway” with previous models but now completes end to end.

These testimonials matter because they describe exactly the kind of reliability gap that has kept many enterprises from moving agentic AI from pilot programs to production deployments. A model that gets 80% of the way through a complex task before stalling creates more problems than it solves; one that reliably completes the full workflow changes the economics of automation. Anthropic also introduced cost-performance curves showing that developers can now adjust effort levels across Sonnet 5 and Opus 4.8 to find the optimal balance of cost and accuracy for their specific use case — a granularity that reflects growing sophistication in how enterprises consume AI services.

An updated tokenizer boosts Sonnet 5 performance but could quietly raise costs for some workloads

One technical detail buried in the announcement’s footnotes deserves attention: Sonnet 5 uses an updated tokenizer that changes how the model processes text, similar to the change Anthropic introduced with Opus 4.7.

The tradeoff is that the same input can map to roughly 1.0 to 1.35 times as many tokens depending on content type. Anthropic says the introductory pricing is calibrated to make the transition “roughly cost-neutral,” but enterprise customers running high-volume workloads will want to benchmark their specific use cases carefully before assuming their bills won’t change.

Anthropic says Sonnet 5 is safer than its predecessor, but its most capable models still lead on alignment

Anthropic’s safety disclosures reveal a nuanced picture. The company reports that Sonnet 5 shows lower rates of hallucination and sycophancy than Sonnet 4.6, is better at refusing malicious requests, and is more resistant to prompt injection attacks in agentic contexts. On Anthropic’s automated behavioral audit — which tests for a wide range of misaligned behaviors including cooperation with misuse and deception — Sonnet 5 scored lower (meaning safer) overall than Sonnet 4.6.

However, Sonnet 5 showed “somewhat higher rates of misaligned behavior” compared with the more capable Opus 4.8 and Anthropic’s Claude Mythos Preview, the company’s powerful but tightly restricted cybersecurity-focused model. On a Firefox 147 exploit development evaluation created in collaboration with Mozilla, neither Sonnet model could develop a working exploit — both scored 0.0% — though Sonnet 5 showed a slightly higher partial success rate (13.2%) than Sonnet 4.6 (8.8%). Both remain far below Opus 4.8 (68.8% working exploits) and Mythos 5 (88.4%).

Because of these incremental gains in cyber-adjacent capabilities, Anthropic launched Sonnet 5 with cyber safeguards enabled by default — real-time systems that detect and block dangerous cybersecurity usage. The safeguards mirror those on Opus 4.7 and 4.8 but are less restrictive than those applied to Fable 5, the latest Mythos-class model that Bloomberg reported on June 10 is “blocked from responding to queries related to cybersecurity and biology.” Organizations enrolled in Anthropic’s Cyber Verification Program automatically receive the same access on Sonnet 5 without needing to reapply.

From $14 billion to $47 billion in revenue: Sonnet 5 arrives as Anthropic’s IPO narrative takes shape

The Sonnet 5 launch arrives at what may be the most consequential moment in Anthropic’s short history. The company confidentially filed its IPO prospectus with the SEC in early June, setting up what CNBC has described as “the most scrutinized public offering in tech history.”

The financial trajectory has been extraordinary. In February, Anthropic raised $30 billion at a $380 billion valuation, with the company reporting $14 billion in annualized revenue that had “grown more than tenfold in each of the past three years,” as The Guardian reported. 

By late May, Anthropic had closed a $65 billion Series H round at a $965 billion post-money valuation — co-led by Altimeter Capital, Sequoia Capital, and others — with a revenue run rate that had crossed $47 billion. Harrison Rolfes, an analyst at PitchBook, told CNBC that the number that will “either validate or collapse the entire narrative the private markets have been pricing for three years” won’t be the valuation or revenue, but gross margin — a figure no outside observer has yet seen.

In this context, Sonnet 5 serves a dual purpose. For developers, it offers genuine capability improvements at competitive prices. For Anthropic’s IPO narrative, it demonstrates the company can deliver a compelling product at a price tier that could drive the kind of broad adoption Wall Street rewards — high-volume, recurring API revenue from thousands of enterprise customers.

Government deals and growing competition define the market Sonnet 5 enters

The timing also aligns with Anthropic’s aggressive push into institutional contracts. Just yesterday, California Governor Gavin Newsom announced a first-of-its-kind partnership providing Claude to all state agencies at a 50% discount, with free workforce training.

Kate Jensen, Anthropic’s Head of Americas, called it an effort to “put Claude to work for the people who keep this state running.” The deal — which extends to California’s cities and counties — represents exactly the kind of durable, recurring adoption that could anchor revenue well beyond the developer community.

But Anthropic’s release lands in an increasingly crowded field. OpenAI, which raised a $122 billion round in March at an $852 billion valuation, is pursuing its own IPO. Elon Musk’s SpaceX, which merged with xAI, priced its IPO at $135 per share with a $1.77 trillion valuation. Google, Meta, and a growing wave of well-funded competitors — including Asian AI startups that, as the Wall Street Journal has reported, are developing Mythos-like cybersecurity capabilities — are all vying for the same enterprise market.

Gil Luria, head of technology research at D.A. Davidson, told CNBC that while Anthropic “appears to have the lead” in frontier AI models, “much of their current usage is for trials and experimentation and that may not sustain.” That observation cuts to the heart of the challenge facing every frontier AI lab: converting experimental developer usage into durable, production-grade revenue.

The real test for Sonnet 5 isn’t benchmarks — it’s whether cheaper AI can sustain a trillion-dollar story

Sonnet 5’s positioning — offering near-Opus performance at Sonnet prices — is a direct play for that conversion. Enterprise customers experimenting with expensive Opus-class models may find that Sonnet 5 delivers sufficient quality for production workloads at a price point that finance teams can approve at scale. If it works, it could accelerate the shift from experimentation to deployment that every AI company needs to justify its valuation.

Three things will determine whether Sonnet 5 matters beyond the initial benchmark charts. Real-world agentic reliability is the first: benchmarks measure capability, but production deployments measure consistency, and the true test will come when thousands of developers push the model through messy, unpredictable workflows at scale.

The tokenizer economics are the second: the updated tokenizer’s 1.0 to 1.35x token expansion could quietly erode the pricing advantage for certain workloads, and enterprise customers should run their own cost analyses rather than relying on headline per-token prices. The third is the IPO narrative itself: when Anthropic’s S-1 eventually becomes public, investors will scrutinize whether the Sonnet tier — cheaper but high-volume — or the Opus tier — expensive but high-margin — drives the bulk of revenue and, critically, gross profit.

As PitchBook’s Rolfes told CNBC, the 2026 IPO window “either becomes the most consequential IPO cycle since the dot-com era or the most expensive lesson in narrative-versus-fundamentals that public markets have ever taught.”

Anthropic is betting that a model good enough to rival its flagship and cheap enough to run at scale is the product that closes the gap between those two outcomes. The public markets will soon decide whether they agree.

Looking to watch CazeTV – the Brazilian YouTube channel with the rights to all 104 World Cup 2026 games? Read on and we’ll show you how to watch CazeTV outside Brazil – including the US, Europe and beyond.

After starting out as a Twitch streamer, Casimiro Miguel has changed the game when it comes to sports broadcasts in the modern era with his CazeTV YouTube channel.

If you like the idea of owning the very best action camera but have so far been put off by the price, then let me bring you some fantastic news. This excellent value GoPro Hero 13 Black bundle is on sale at Amazon for $379.99 (was $479.99).

That’s a new record-low price for the highly rated 5.3K resolution camera from the industry’s leading brand. Considering the camera ordinarily retails for $429 by itself, this bundle deal is even more impressive.

In addition to the excellent GoPro camera, the bundle features a handle, two Enduro batteries, two curved adhesive mounts, a 64GB SanDisk MicroSD card, and a handy carrying case. That’s everything needed to get filming as soon as it arrives at your door.

Competition for the top spot in our best action camera guide is fierce, but the GoPro Hero 13 Black has held prime position for a long time now. Its commendation is in large part thanks to the excellent-quality 5.3K video, a superb range of accessories, including new auto-detected Lens Mods, and improved battery life and heat dissipation.

You can read more about why we love it in our GoPro Hero 13 Black review. Specifically, we praised its Bluetooth audio support, versatile mounting options, and ability to capture great-looking footage in well-lit areas.

I don’t record much video for social media specifically, but if you do, then you’ll love the 27MP sensor, which enables footage to be recorded in an 8:7 aspect ratio. Say goodbye to the awkward cropping of landscape footage for portrait outputs.

Other Slushie Machines I Liked

Photograph: Matthew Korfhage

GreenPan

Frost Slushie Machine

The slushie machine from Belgian-founded kitchen and wellness brand GreenPan is maybe the only slushie machine I’d describe as being even slightly attractive, or pleasant on a countertop—available in a trendy pistachio color scheme that a 21-year-old co-tester called “cute.” The slush produced by this device also had quite a nice consistency, perhaps due to a tighter auger around the cylinder that roiled the slush a little more. My colleague Martin Cizmar, who also tested this device, was able to recreate a Philly recipe for Italian-style water ice with Meyer lemons, and declared himself an unending fan.

The GreenPan slushed admirably, making a full chamber’s worth of spiked slush in about 25 minutes. This is nowhere near as fast as the XL or the Twist on slushing speeds, alas. The fill chamber is a little shallow, which means you have to pour slowly or you’ll make a mess. If you accidentally leave the handle down, you’ll also make a mess. Some reports online of cracks in the cylinder over use are also reason for pause. But if aesthetics are a prime consideration, this will slush handily. And look better while doing it.

The original Ninja Slushi was quite simply a triumph of industrial design when it arrived in 2024—the machine that managed to bring the cocktail bar or convenience-store slushie to the home kitchen countertop. Among many imitators, Ninja’s original design remained the most user-friendly and reliable until the next-generation Ninjas supplanted it.

I’ve made coconut-lime daiquiris for a family of visiting Brazilians, who joked that they planned to take the machine back with them on the airplane. I’ve entertained a party full of children with the nonalcoholic version of slushie. And I’ve made silly frozen cocktails at home, whether lime Jarritos slushies or tamarind michelada slushies. Everything frozen is better, it turns out. Freezing a cocktail adds fun and removes shame.

But it’s been replaced. I consider the original Slushi a good value model, but it’s no longer the top of the market. The original Slushi doesn’t slush as well on higher-alcohol slushies as the newer XL and Twist, even for ABV below 16 percent. (Really, with an OG Ninja Slushi, the sweet spot is around 10 to 12 percent ABV if you want good consistency.) Milkshakes/soft-serve are not really feasible on the original Ninja either, always either foamy or ice-gritty.

Which is all to say, buy the Slushi when it’s on a good sale at $250 or less—it served me well for a year—or when it’s updated with a compressor as good as the one on the XL or Twist.

Other Slushie Machines Tested

Ever since Ninja took slushies to the home market, the Amazon directories have filled with newer brands you’ve likely never heard of and whose names sometimes seem subject to a randomizer engine: Inoviva, Chivalz, Vibofrost, Friwest, Aekda, Syintao, Vischic, Ranvaira, Rinvotio, and the list goes on. Most are available at discounts compared to Ninja or other more recognizable brands.

I’ve tested three such brands: Chivalz, Invoviva, and Vibofrost. All three have had one form of reliability issue or another: basic design defects, inconsistency of performance, or simply disappearing from the market.

Chivalz Slushie Machine (no longer in stock): This was previously WIRED’s budget pick, which my co-tester Kat Merck called, without insult, “a quite respectable Ninja Slushi knockoff.” The device arrived with a welcome digital temp readout and a removable back panel that made cleaning easier on the slush chamber. Performance was comparable to the original Ninja, though the user interface was a bit janky. But since last year, the brand’s slushie machines seem to have disappeared, as the brand’s focus returned to air purifiers and humidifiers.

Vibofrost Slushie Machine ($235, sold out after Prime Day): This Vibofrost, like the Chivalz, freezes slushies comparably to the original Ninja Slushi. And like the Chivalz, it has a somewhat irritating child-lock feature, and a timed feature that seems of limited utility. Though it will slush within around 20 to 30 minutes, the oddly designed spout can spray wildly if there’s any liquid in the machine, the drip tray does not attach securely, and it kinda moans like a dying tauntaun while in operation.

Inoviva Slushie Machine for $120: I tested this Inoviva slushie machine twice. The first time, the device registered much louder than competitors, the drip tray arrived stuck to the machine, and the compressor began to fail after a week’s testing. The second time, it was still loud, and the user interface had a difficult-to-navigate locking feature, but freezing was indeed more consistent. The inconsistency in quality control makes this device difficult to recommend. But maybe you’re willing to brave this for a steeply discounted price. The Inoviva also has one terrific feature: The ability to adjust thickness for each drink setting.

AccordionItemContainerButton

My co-tester Kat Merck (on the now-discontinued Chivalz) and I made so very many slushies with each machine, from dairy to nondairy to coffee slushies to straight-up bottles of wine. Specifically, we tested every version of slush that a machine advertised. If Ninja or GreenPan says a machine can make frappés and milkshakes and frozen juices, we made frappés and milkshakes and frozen juices, tinkering where necessary. I froze orange juice and strawberry juice, slushed a bouquet’s worth of rosé, and made slushies from daiquiri to margarita to whiskey Coke. I slushed tamarind micheladas (an excellent idea) and Twisted Tea (a terrible idea).

Photograph: Kat Merck

I also raced the freezing capabilities of each machine by pouring a 16-ounce can of delicious Mike’s Harder Lemonade in each, then seeing which machine was fastest. (For the XL, I used a 24-ounce can.) And I made smooth and dense coconut-lime daiquiris with coconut milk, according to Ninja’s recipe, to test how well each machine’s dispenser handled a genuine dense-textured challenge.

How Do Home Slushie Machines Work?

AccordionItemContainerButton

The tech is pretty simple, almost ingeniously so: A beefy cylindrical freezing core in the center of the drink chamber continually cools any liquid in contact with it. It’s encircled by a plastic spiral auger attached to a motor. The auger mixes the drink, keeps it slushing instead of freezing solid, and also pushes the resulting slush toward the dispenser nozzle so you can have some. The resolute simplicity of this design allowed Ninja and others to scale down the commercial slushie maker for home consumers thirsty for frozen treats.

Advertisement

The main requirement on most machines is that the frozen beverage have more than 4 percent sugar—or between 3 percent and 16 percent alcohol (20 percent for the newest Ninjas). This lowers the freezing point of the resulting concoction, and makes slushing possible. Some slushie machine vendors recommend percentages more like 15 percent sugar, for perfect consistency. But I often balk at this. Coca-Cola and orange juice are each around 11 percent sugar—so that’s very sweet. Some hero of the internet has made a slush calculator for easy reference.

A minimum of 16 ounces of liquid is required for most 88-ounce home machines, for simple reasons: The liquid needs to be in physical contact with the core in order to slush up and also to keep ice from forming on the central cylinder’s surface. The Slushi XL requires a 24-ounce minimum, because it’s bigger.

Can You Put Diet Soda in a Slushie Machine?

AccordionItemContainerButton

No and yes. Slushies rely on a helpful property of water: Sugar (or salt) dissolved in water lowers its freezing point below 32 degrees Fahrenheit. Why? Solubles like sugar are chaos agents. Sugar molecules move randomly, refuse to dissolve into ice, and interfere with water’s ability to form hydrogen bonds and turn crystalline. Some water molecules freeze, but sugar water doesn’t. Tada! Slush.

Advertisement

If you try to make a slushie out of sugar-free soda, or sugar-free anything, ice crystals will instead form easily. The stainless steel freezing core will ice over and scrape on the auger, and ice cubes or hunks will gather mass in the slushie machine. The cylinder will start to shake, then the machine will clunk, then eventually you’ll probably break your machine: Low-sugar fail-safes on these devices have not been overly reliable, alas. So don’t try this at home!

This doesn’t mean you’re doomed to massive calories if you want to make a slushie. Not every artificial sweetener lowers the freezing point appropriately, but the one that Ninja recommends for diet slushies is allulose, a rare but naturally occurring sugar that’s 70 percent as sweet as basic sugar but is not metabolized effectively by the human digestive system. This means it’s low in calories and doesn’t cause insulin spikes—but as with a lot of indigestibles, note that side effects can include bloating or GI distress for some.

For easiest use in a slushie, buy liquid allulose. Powdered versions also exist, but to use them, you’ll need to make a simple syrup by heating up the powder in water to help it dissolve, then let it cool. If you just try to drop the allulose powder into your machine with some Diet Coke, it might not dissolve, and you might still get ice formation. Or at least, I definitely still got ice formation when I tried this on the OG Ninja, and had to stop my machine.

How Can You Stop Milkshakes From Getting Foamy in a Slushie Machine?

Advertisement

AccordionItemContainerButton

Bet you didn’t expect a lesson in milk proteins today! But here’s the deal: Milk proteins start to separate when agitated. Churning milk is, in fact, how butter gets made. Proteins separate out, and you get butter on the one side and buttermilk on the other. Both are delicious, but neither is wanted in a milkshake.

Photograph: Matthew Korfhage

If you try to make a milkshake in a churning slushie machine using just milk, you’ll eventually start to see the effects of these milk proteins separating out from buttermilk—which will manifest first as an undesirable foaminess. To avoid this, Ninja recommends also adding heavy cream or half-and-half to any milkshake recipe. The higher fat content will keep things smoother.

Note it’s easiest to use fruit syrups, rather than just juice, and add vanillin, or it’ll be a bit boring: The heavy fat tends to overwhelm any subtle fruit flavors. Another deep secret of the tasty milkshake? Salt. Add a tiiiinny pinch; it’ll help bring out flavor. A 16-ounce McDonald’s milkshake has 260 milligrams of sodium—about 1/16th of a teaspoon of table salt, or approximately the amount that fits between your index finger and your thumb.

But temper your expectations here. None of the slushie machines we tested made a texture comparable to a classic milkshake. On most machines, which don’t have compressors as powerful as the new-model Ninjas that are now our top picks, the texture is often a little ice-gritty and not as richly textured or integrated as the milkshake you’ll get from your local burger joint, let alone the soft serve from the famously broken ice cream machines at McDonald’s. Slushie machines also can’t handle chunks of frozen fruit, often the best part of a milkshake.

Advertisement

On the newer Ninjas, with their more powerful freezing power, I was able to get the smoothness and freeze I wanted. But because most recipes call for a high-fat mix of 2:1 milk and heavy cream in order to avoid churning foam and butter, the results still weren’t quite a light milkshake. It was more like a dense, rich, quite tasty soft serve. I learned the hard way that throwing in a half-teaspoon to a teaspoon of salt was necessary to drop freezing temps enough to get good cream formation.

Now, do I like being able to make 20-minute soft-serve in my home? From milk and heavy cream and sugar and a dash of salt and vanilla? Heck yes, I do.

Dirac has renamed its PC and Mac-based Dirac Live Room Correction Suite as the Dirac Live Processor, expanding the software platform to include Dirac Live Active Room Treatment (ART). The change gives enthusiasts, home-theater owners, and audio professionals a computer-based route into Dirac’s full room-optimization ecosystem, including Dirac Live Room Correction, Bass Control, and ART, without replacing their existing source hardware.

That does not mean every networked audio system suddenly becomes ART-capable. The Dirac Live Processor can serve as the software hub for compatible playback systems with the appropriate licenses, and a properly measured speaker configuration. For listeners already using a PC or Mac at the center of a serious audio system, however, it creates a far more flexible path to Dirac’s most advanced low-frequency and room-acoustics tools.

What Is the Dirac Live Processor and Who Is It For?

The Dirac Live Processor is a virtual audio processor for PC and Mac that applies room correction to audio before it reaches the sound system. It gives listeners a way to measure and optimize room and system performance using software and the computer they already own, including with audio systems where Dirac Live is not built in.

With continued support for VST, VST3, AAX, and AU plug-in formats, the Dirac Live Processor can also be used with compatible DAWs (digital audio workstations) and media players as part of an existing computer-based playback setup, including studio systems.

The Dirac Live Processor incorporates Dirac Live Room Correction, Dirac Live Bass Control, and Dirac Live Active Room Treatment. Together, these technologies establish the Dirac Live Processor as Dirac’s PC and Mac platform, giving listeners a single place to access its full suite of room-acoustics tools.

“Active Room Treatment is the most advanced room acoustics work we’ve done, and Dirac Live Processor is how we bring it to PCs and Macs,” said Nilo Casimiro Ericsson, Product Manager for Dirac Live. “Starting today, anyone can install it on their PC or Mac and hear the difference Dirac Live Active Room Treatment can make in their own sound system, in their own room.”

Dirac Live Room Correction analyzes how a room and speaker setup affect the interaction between sound and space, then applies corrections intended to improve timing, phase alignment, frequency response, imaging, and tonal balance.

Dirac Live Bass Control optimizes low-frequency performance across speakers and subwoofers, aiming to deliver smoother, more consistent bass throughout the listening space.

With the addition of Active Room Treatment, the Dirac Live Processor supports Dirac’s complete approach to room acoustics management.  Active Room Treatment builds on Dirac Live Room Correction by using the full speaker array as a coordinated acoustic system, actively helping control room resonances and sound decay to deliver a cleaner, more controlled soundstage with greater clarity, detail, and focus.

Over the past 12 months, Dirac has expanded access to Dirac Live Active Room Treatment through a growing range of home-audio collaborations and integrations with brands including AudioControl, Denon, Marantz, miniDSP, Monoprice, and StormAudio.

The Dirac Live Processor extends that access to PC and Mac users who want a software-based approach to room-acoustics optimization, without requiring their AVR, preamplifier, or other system component to have Dirac technology built in. Setup begins by connecting a measurement microphone to the computer. The software then guides users through the room-measurement process, analyzes room and system behavior, and creates filters tailored to the listening environment.

“With Active Room Treatment now available, Dirac Live Processor becomes a powerful way to experience our most advanced room acoustics technology in any system,” continued Casimiro Ericsson.

Pro Tip: To use the Dirac Live Processor with an AVR or related component, the device and the PC or Mac running the software must be connected to the same network.

Dirac Live Processor Quick Start Guide

The Bottom Line 

Dirac Live ART has quickly emerged as one of the more sophisticated room-acoustics solutions available to home-theater and two-channel listeners. With its addition to the newly renamed Dirac Live Processor, PC and Mac users can now access Dirac Live Room Correction, Dirac Live Bass Control, and Dirac Live Active Room Treatment from a single software platform.

Three Key Points

• Dirac Live Room Correction Suite has been renamed Dirac Live Processor.
• Dirac Live Active Room Treatment has been added to the Dirac Live Processor platform.
• A PC or Mac can serve as the control and processing hub for Dirac Live in systems without Dirac built into the source hardware, although compatibility, licensing, and network requirements still apply.

There is no free lunch in acoustics, unfortunately. Unlike the basic room-correction systems bundled into many AVRs, Dirac Live Processor features require separate user licenses, with pricing dependent on the level of correction and bass-management capability required.

The Dirac Live Processor is aimed at serious computer-audio users, home-theater owners who may not have Dirac-compatible hardware (however, network connectivity is required), and studio or enthusiast listeners willing to measure their rooms rather than simply hope the sofa is in the right place.

When properly implemented, Room Correction, Bass Control, and ART can improve bass consistency, imaging, tonal balance, and overall clarity. Casual listeners may struggle to justify the added cost and setup, but for anyone trying to extract the full potential from a good loudspeaker system in a less-than-perfect room, it is a meaningful and potentially transformative tool.

Pricing & Availability

Active Room Treatment is available on the Dirac Live Processor starting June 30, 2026. Existing users of Dirac Live Room Correction Suite will automatically be transitioned to Dirac Live Processor together with their current licenses and settings. Licenses can be purchased with an optional 14-day free trial at www.dirac.com.

• Dirac Live Room Correction – $349 (for Mono and Stereo) $499 (for Multichannel)
• Dirac Live Bass Control – $299
• Dirac Live Active Room Treatment – $299

Pro Tip: Dirac has a dedicated landing page for Dirac Live Processor licenses. During the launch period, this will be the main place to buy ART/DLP on the site. Going forward, Dirac Live Processor will be added to Dirac’s main device list when selecting a license, so that it shows up as an alternative alongside compatible hardware devices.

Related Reading:

“The more we scaled, the more we burned”

It’s been a long road for WhyQ.

In 2016, the company was built around a straightforward idea: make hawker food more accessible to busy office workers in the CBD. After years of pivots, a pandemic-era detour into residential deliveries, and the hard lessons that came with it, the Singapore startup has appeared to have found a model that works—B2B corporate dining.

And the numbers are starting to show it.

WhyQ co-founders Varun Saraf and Rishabh Singhvi shared that the company hit baseline profitability in Singapore in Q2 2025 and has stayed there. They added that the business has maintained an annualised revenue run rate of about S$5 million through Q1 2026.

For a company that was burning through cash during its pandemic-era consumer delivery phase—dispatching individual riders for meals worth S$10 to S$15 with no economies of scale—that’s a meaningful turnaround.

“The more we scaled, the more we burned,” COO Singhvi said in a previous interview with Tech in Asia. The pivot to B2B, he shared, “solved our entire unit economics puzzle.”

Delivering over 2,500 corporate meals daily

WhyQ now operates exclusively as a B2B platform. Gone is the sprawling consumer app that once partnered with over 2,200 hawker stalls across 35 hawker centres.

In its place is a leaner corporate dining operation: just slightly over one-fourth the number of partners at 500 merchant partners, but with long-term contracts that give the business predictable, recurring revenue.

The biggest lesson [we] learned was trying to compete in the volatile consumer delivery market, where revenue simply lacks long-term predictability.

About 20% of orders still come from hawker partners, with the remaining 80% driven by curated restaurant brands. Meal prices range from S$8 to S$25 per head.

The company has also moved away from the third-party gig logistics that caused so many of its consumer-era headaches. WhyQ now runs its own dedicated delivery fleet, whose riders are trained to conduct quality checks at the kitchen and handle meal setup directly at client offices.

“HR doesn’t have to lift a finger,” Saraf said.

Today, WhyQ delivers more than 2,500 meals daily to corporate clients.

What makes it even more appealing to corporate clients is that WhyQ white-labels its ordering portals so they look like the client’s own internal platform. It also handles everything from daily lunches to pantry snacks, event catering, and live food stations.

The company claims 100% client retention among its corporate accounts.

Merchant partners are also benefiting from the predictability. Gyoza San’s founder Wilman Ng said the corporate order program has consistently boosted their monthly revenue by 15 to 20%. KinBaba Thai reported a similar 15% revenue lift since joining WhyQ’s network.

WhyQ’s next big bet

Assuming the B2B foundation holds, WhyQ’s next move is WhyQ Intelligence, an AI-powered nutrition and wellness tool currently in pilot, with a commercial launch targeted for Q3 2026.

The idea is to let employees track nutritional macros, set dietary targets, and chat with an AI assistant for menu recommendations—all tied to their company’s curated meal options. For HR teams, it connects meal participation data with attendance trends and employee engagement metrics.

Saraf frames it as both a client retention play and a new data layer: understanding what employees actually want to eat, which keeps menus fresh and deepens WhyQ’s grip on the accounts it already has.

The company is targeting a 70 to 80% engagement rate among employees at client companies in the tool’s first year.

Down the line, WhyQ Intelligence could potentially be spun out as a standalone product sold to companies managing their own food programs. That’s a longer runway, but it signals the founders are thinking beyond logistics.

“We have a long way to go here”

WhyQ is targeting 50% year-on-year revenue growth, driven largely by expanding within existing accounts. One enterprise client is reportedly adding 120 daily meals in Jun 2026 and another 250 in Oct.

Beyond Singapore, the founders see Hong Kong and Sydney as their next potential markets, pointing to similar competition for talent and existing enterprise customers with offices in both cities. Rather than expanding speculatively, WhyQ says it will only enter a new market after securing a profitable anchor contract.

It also plans to acquire local merchant networks where possible instead of building operations from scratch.

That said, overseas expansion isn’t new territory for the startup.

WhyQ previously operated in Malaysia, where it offered two free digital tools: WhyQ EBiz, an app that helped merchants manage their businesses online, and WhyQ Kira Kira, a digital bookkeeping app. The company has since exited the market.

We chose to exit that market because we feel we are still just scratching the surface of the Singapore market and have a long way to go here.

That opportunity remains significant.

Over the next three to five years, the founders plan to deepen their reach into industrial and commercial food deserts like Tuas and Jurong, areas where workers have historically had limited access to quality, convenient food options.

Years of pivots, a pandemic, and painful lessons have reshaped WhyQ into a business focused less on growth for growth’s sake and more on sustainable economics.

Whether that formula holds remains to be seen. But for now, the startup appears to have found something it spent nearly a decade searching for: a business model that actually works.

• Read other articles we’ve written on Singaporean businesses here.

Also Read:⁠ This 52 Y/O kopi business roasts 1,000kg of coffee every month & is winning over younger drinkers

Featured Image Credit: WhyQ