With an emerald and gold finish, the Zelda 40th Anniversary Switch 2 finally gives us a fun-looking Switch 2 console. and it makes standard black look drab by comparison.

As well as the new console, there will be a Zelda anniversary edition Switch 2 Pro Controller, available with and without a display stand, and there's also a case with screen protector. All of them feature a golden Triforce emblem as a design centerpiece.

There are some lovely details, like the gold accents around the thumbsticks and the subtle triangular motifs, making these truly look like collector-worthy editions.

Dedicated Zelda fans have reacted positively, say the console feels like a proper tribute to the franchise’s 40th anniversary this year, although a few worry about covering up the dock’s front emblem with third‑party accessories, which would hide the standout design.

The big question is how much will it cost. The standard Switch 2 retails at $499 in the US following a price hike earlier in the year (thanks, Donald). Some fans fear the special-edition will get a further mark up, but Nintendo hasn't confirmed that. Pre-orders have yet to open, but keep an eye on the Nintendo store.

Yuta ONE is very active on social media, sharing work updates and tips, and he's been offering free fonts to attract more followers. The latest giveway of around 200 OTF files is to celebrate the artist reaching 170,000 followers on X, but you'll need to be quick because it runs until 9 September until midnight in Japan.

Yuta ONE discovered street art as an inline skater and began teaching himself graffiti and calligraphy and later typography – mainly black letter typefaces. He's known for artwork such as the title logo and jacket illustration for Tokoyami Towa's SM4SH IT UP single.

The free font pack covers a range of styles, including futuristic looks as well as graffiti and gothic fonts. You can download it free font pack via Booth here.

Yuta ONE says that the giveaway will be extended, with another 11 unpublished fonts added, if the follower count reaches 190,000 before it lapses. There's also talk of a photo in a micro bikini at 200K.

An AI laptop lets you take advantage of today's AI-assisted creative tools which run locally rather than in the cloud (denoising, upscaling, auto-rotoscoping, motion tracking), without draining the energy of the CPU and GPU. If you already have our list of laptops for graphic design bookmarked, this is the specialised, more demanding tier above it.

Our overall pick for the best AI laptop for motion design and VFX is the Apple MacBook Pro 16-inch (M5 Max, 2026), thanks to its combo of raw multi-core power, a hugely capable GPU and Apple's most advanced Neural Engine yet. We've also included alternatives if you're specifically looking for a Windows laptop or a workstation, while there's also a budget option if you're funds are more limited.

Best overall

If you're running demanding motion graphics or VFX software, the more powerful the machine, the faster and smoother everything will go. And the MacBook Pro with M5 Max is built for exactly this kind of work.

Its 18-core CPU pairs Performance and Efficiency cores with a beefed-up GPU, and each GPU core now carries its own Neural Accelerator, so AI-assisted tasks in Final Cut Pro, DaVinci Resolve and Cinema 4D can run locally without waiting on the cloud. As a result, VFX previews render quickly, with graphics performance up around 50% on the previous M4 Pro and M4 Max generation, and DaVinci Resolve Studio rendering as much as five times faster.

With up to 128GB of unified memory and 4TB of storage, this super-powerful laptop is able to deal with enormous project files without complaint, and there's also excellent battery life alongside all that power. It is, however, very expensive: a fully-specced M5 Max configuration runs into five figures; something you might find that galling, given that Apple still hasn't brought OLED or a touchscreen to the Pro line. For anyone serious about motion graphics or VFX who can stomach the cost, though, this may well be an investment in time and efficiency that pays off in the long term.

Read our full review of the MacBook Pro with M5 Max

Best Windows AI laptop

If you want a Windows laptop, the ASUS ProArt P16 is the closest thing to a MacBook Pro rival built specifically for motion work. Its RTX 5070 GPU marks a huge step up from the previous model, offering a 25-28% jump in graphics performance, and Premiere Pro, Photoshop and generative AI tools all run smoothly even with feature-heavy files. Its AMD XDNA NPU, rated up to 50 TOPS, also handles a lot of the on-device AI legwork in ASUS' own Creator Hub tools, keeping tasks like sketch enhancement and image generation off the GPU so it's free for rendering.

The 3K OLED touchscreen, meanwhile, measured close to full DCI-P3 coverage with excellent colour accuracy, making it a strong option for artists who want hands-on, stylus-driven control over their work. (Apple, let's not forget, still hasn't given us a touchscreen.)

The trade-offs are noise and heat under sustained rendering loads, plus a proprietary charging port that locks you into ASUS' own cables. Plus, as you might expect, it's pretty darned expensive. With all that in mind, though, we still rate this as the best all-round Windows AI latop for creatives working in motion graphics and VFX today.

Read our full review of the ASUS ProArt P16.

Best budget option

For motion design and VFX professionals on a tighter budget, the ASUS TUF Gaming A16 proves that even an entry-level discrete GPU makes a big difference over relying on a CPU's integrated graphics. Its RTX 5050 handles Photoshop and everyday 2D creative work comfortably, and it isn't bad at 3D and AI-assisted apps either, even if it drops behind pricier machines once you move into heavier video editing in Premiere Pro. Its NPU is modest next to the other machines here though, rated at just 16 TOPS, so it's the GPU rather than on-device AI that's doing most of the heavy lifting.

The screen is bright and covers sRGB in full but falls short of the wider DCI-P3 and Adobe RGB gamuts that serious colour work benefits from. For that reason, this laptop is better suited for layout, compositing and general motion graphics than colour-critical grading. Battery life, meanwhile, is a surprise for a gaming laptop, delivering nearly eight hours in everyday use. It's not the best AI laptop for motion graphics and VFX work, but it is an affordable entry point when a full workstation-grade GPU is out of reach.

Read our full review of the ASUS TUF Gaming A16.

Best workstation

Value power over portability? The Legion Pro 7 is the closest thing here to a desktop workstation you can close and carry. Its RTX 5090 GPU comfortably outpaces every other laptop on this list for rendering and 3D work, and it can demanding both creative and gaming workloads without breaking a sweat.

Its Intel AI Boost NPU plays a smaller role than the GPU in day-to-day use here, but it still helps to offload lighter background AI tasks. The 16-inch OLED display covers the full DCI-P3 colour space, so it's accurate enough for serious photo and colour work alongside VFX previews.

The compromises are the ones you'd expect from a machine this powerful: it's bulky, heavy, and battery life away from the mains is limited, especially once the GPU is under load. There's also no SD card slot, an odd omission for a laptop aimed partly at creatives. But for studios or freelancers who need workstation-class performance in a portable shell, it's our top recommendation.

Read our full review of the Lenovo Legion 7 Pro.

FAQs

And, well, it looks like Nintendo has been busy. The remake is getting a pretty serious visual overhaul, with Hyrule rebuilt in much greater detail, Link looking considerably more expressive, and the whole thing sitting somewhere between the familiar look of Ocarina of Time and the more polished fantasy style we've come to expect from modern Zelda games, which should put to rest any feelings that this remake won't meet expectations.

The big thing is that Nintendo doesn't appear to be treating this as a quick lick of paint. The environments have been rebuilt with modern materials and lighting; there's a free camera, and Link has more freedom of movement, including jumping, diving, and rolling around Hyrule rather than relying entirely on the movement rules that worked on the N64.

The Ocarina remake feels ambitious

This approach makes sense because Ocarina of Time is now 28 years old, and while its design still holds up remarkably well, some of the elements around it definitely belong to another era. Nintendo is being careful about how far it takes those changes, though. The remake is still very much Ocarina of Time, rather than Nintendo taking the original and trying to turn it into another Breath of the Wild.

During Nintendo's Zelda 40th Anniversary Direct, Eiji Aonuma described it as a "faithful interpretation using modern technologies", which is the right approach for a game this beloved. Change too little, and you end up with something that feels like an expensive texture pack; change too much, and somebody, somewhere, will be outside Nintendo HQ with a very angry sign.

More interesting changes are tucked away in there too. The ocarina, for example, gets a new modern twist. Instead of simply remembering the button combinations for each song, Switch 2 players can apparently hum, sing or play the notes into the console's microphone, with Link then playing the tune if he knows it. Nintendo is also adding a "threads of time" system to help with navigation, while time itself continues to pass as you explore towns and the wider world.

Then there's Link himself, who has had the sort of glow-up you'd expect after nearly three decades. The new model has much more detail in his clothing, face, and equipment, but Nintendo has kept the familiar silhouette and overall character design intact, which will probably matter more to fans than whether every individual texture is technically impressive.

That's the tightrope Nintendo is walking with the Ocarina of Time remake, because people don't necessarily remember what the N64 version actually looked like. We remember how it felt, and that emotional connection shapes what we imagine it looked like.

For anyone who played it in 1998, that blocky little version of Hyrule has probably been receiving a free internal remaster every time you think about it. Your brain fills in the missing detail, the lighting, the scale, the atmosphere, all of it. So Nintendo isn't really competing with a 28-year-old game here; it's competing with people's memories of it.

The first teaser for the remake, shown back in June, gave us only a brief look at young Link, so this is the first time we've really seen the broader visual direction. And from what Nintendo has shown, it looks considerably more ambitious than a simple HD conversion.

Whether all those changes work once we're actually holding the controller is another question, but at least Nintendo seems to understand that remaking Ocarina of Time isn't about making everything shiny and calling it a day. It's about figuring out which bits need dragging into 2026 and which bits you absolutely leave alone.

We'll find out when The Legend of Zelda: Ocarina of Time arrives on Nintendo Switch 2 on November 5.

But although Blender can be used for texturing, most 3D artists I speak to use Adobe Substance 3D Painter. It's faster, more specialised and better integrated into professional pipelines. However, a new Substance 3D Painter alternative is showing a lot of promise. Open-source ArmorPaint is lightweight and now available for whatever platform you might want to work on.

Developed by Lubos Lenco and the team behind the open-source game engine Armory3D, ArmorPaint has been in early access for eight years. Version 1.0 has finally been released into the wild, and it's available to download for Windows, Linux, macOS, iOS and Android (the latter two being experimental for now).

The app allows real-time texture painting in a path-traced 3D viewport and can be used to generate PBR texture maps ready for use in the best 3D modelling software, animation software or game engines as PNG, JPEG or EXR at up to 32-bit colour depth. Assets can be brought in via the usual formats or as Blender scene files, and materials can be created via a node-based material editor or by importing PBR texture maps.

The app has a custom-built 3D engine and 2D drawing system and supports ray tracing via Direct3D 12, Vulkan, Metal, and WebGPU graphics interfaces. Recent updates have added Decimate, Smooth, Bevel, and Subdivide modifiers for mesh editing, a basic UV map editor, Neural Nodes for generative AI features, photo-to-material generation, skinning animation playback and support for tween animations.

A code rewrite in C allows interactive painting on bigger 3D assets, with more texture layers at higher resolution. A clean up of the codebase has also slashed the program's size. Armory's team says it could now fit on two 1.44MB floppy disks.

On desktop, it runs as a sub- 10MB portable application with no installation and near-instant start up (you might want one of the best laptops for 3D modelling or the best graphics cards). There are also plugins for Blender, Unity and Unreal Engine.

See the videos below for walk through on some of the recent updates. You can download the free texture painting software from ArmorPaint.org.

In many ways it's a very 'digital' style, and one I've drawn inspiration from when using the best digital art software. Musil works quickly, does relatively little blending, builds his paintings up in layers, and lets the brushwork remain part of the finished image – Musil has talked about his process in more detail on his website.

This is a useful workflow for digital artists to ‘steal’ because we've got an almost ridiculous number of ways to smooth, correct and polish everything… so let's do the opposite; let's try and keep things simple. It's something I do when reviewing the best drawing tablets: to test and have fun. Below, I go into a little detail on Musil's work and how you can draw inspiration from it (he also runs classes).

01. Don't start on white

Musil uses a warm orange-gold ground as the starting point for his paintings, using Quinacridone/Nickel Azo Gold mixed with glazing medium. You can adapt that idea digitally in about ten seconds by creating a layer underneath everything else and filling it with a warm ochre, orange or reddish brown. You don't need to cover it completely later, either; let little bits peek through between your strokes. This is lovely to do in Procreate (read our Procreate tutorials), or even in ArtRage using the Roller, which can pick out the grain in the digital paper, and it's easy to use to create a texture to paint over.

Sometimes in Rebelle I'll add paint directly into this 'wash' layer, using a dirty brush and letting the colours mix live and see what happens. I love to keep all the mess and abstraction in digital paintings, largely because I'm doing this for myself and having fun; there's no commercial goal.

It gives the whole painting a shared warmth and, perhaps more importantly, gets rid of that slightly clinical feeling you can get when every digital painting starts life on a perfectly white canvas.

02. Start with fewer colours

One interesting thing about Musil's process is how little he seems to need in terms of palette. He's talked about moving away from buying lots of different colours and instead mixing what he needs, even mentioning in the Ampersand blog how all those unused green paints sit around because he prefers mixing his greens from yellow and black.

You can replicate that idea very easily in Photoshop, Rebelle or Procreate – read our Procreate art secrets guide to how to easily set up a limited palette. Before you start, choose maybe five or six colours and stick with them. Pick a dark, a light, a warm colour, a cool colour, an accent, and something somewhere in the middle. If you're using reference photos, you can turn the colours into a simplified palette.

Then make the rest by changing the value and temperature of those colours rather than opening the picker every 30 seconds. It sounds restrictive, and it is, which is exactly why it's useful. You'll probably find that the colours start balancing and working with each other much more naturally.

03. Stop blending everything

Digital painting makes blending very easy, so we tend to do it whether the painting needs it or not. You paint two colours next to each other, grab a soft brush, smooth the transition, zoom in, smooth it again, and keep at it until all the personality is wiped away, or at least I do.

Musil's process is the opposite: he starts with big strokes and broad areas, then gradually works toward the details, rather than trying to make every transition perfectly smooth from the beginning.

Grab a reasonably opaque brush, make the sky with a handful of strokes, block in your trees as shapes, throw the highlights over the top, and resist the urge to polish every join and blend edges. Then zoom out to 25% or 50%; all those slightly scruffy marks start to work together.

04. Make your brush strokes meaningful

Ensuring each brush mark means something is another good habit to adopt. Don't think of a brush stroke as simply the thing that puts paint on the canvas, and instead think about what the stroke is doing. A long horizontal mark can describe water, while a vertical one can help sell a tree trunk, and diagonal strokes can lead the eye into a landscape. Short, broken marks can suggest foliage without forcing you to paint every leaf.

Musil's brushwork has this sense of direction, so the marks aren't simply texture sprinkled over a finished painting; they're helping construct the image. You’ll want a good drawing tablet to get the most out of this technique, as you’ll ideally need to adjust the brush mark by rolling the stylus in your fingers (Apple Pencil Pro and the best iPad for drawing are great here), but you don't need to download some enormous 300-brush pack to achieve Musil's look and feel.

A basic textured, opaque brush will do; then change the size, pressure, and direction, and let the marks vary. The important bit is resisting the urge to tidy them all afterwards. Also try using layers, and build up blocks of brushstrokes and colour from dark to light.

Try it yourself: give yourself 10 minutes and five colours.

Fill the canvas with a warm ground, block out the biggest shapes with a large brush, add another layer of chunky strokes, change the direction of those strokes to follow the forms and then, right at the end, add a handful of smaller marks.

Now stop, and don't do the usual thing of zooming in looking for problems. Don't start cleaning edges and don't add 14 new colours because you suddenly decide the sky needs a slightly different blue.

Zoom out and look at the painting as a whole. If the big shapes work, the colours hang together, and your brush marks have a bit of energy, you've probably got a more interesting image than you would after three hours of smoothing and fiddling.

With digital art software, we have endless brushes, infinite colours, and layers you can edit and tweak, so giving yourself fewer choices can actually make painting more interesting. That's been my overall take after looking at Musil's approach and adapting parts of it to digital painting.

So when NVIDIA demoed a local AI agent running on an RTX Spark‑powered laptop — pitched explicitly as your personal chief of staff — the promise landed squarely in that pain point. A machine that reads your inbox, understands your calendar, pulls your decks, tracks your projects, and tells you what actually matters today? That’s not a productivity tool. That’s a fantasy.

But the demo showed something more interesting: not the fantasy, but the tension between what’s possible, what’s plausible, and what’s still very much PowerPoint‑stage ambition.

The pitch: your workday, decoded by a local model

The presenter started with a familiar scene: the morning email flood, the overlapping meetings, the multitasking spiral. Then came the pitch: an AI agent running locally on an RTX Spark laptop, with full access to your Google Workspace. Calendar, Gmail, Drive, Slides, Sheets, Docs. All of it.

The agent’s job?

Tell you your top priorities. Summarise what’s done and what’s next. Pull the right files. Read the right threads. Offer to take work off your plate.

In other words: the assistant that actually does the assistant part.

And to its credit, the demo made that vision feel tangible. The agent pulled an executive‑review deck, scanned email threads, and started assembling a concise briefing. It even offered follow‑up actions; the sort of nudges that normally require a human who’s been paying attention.

This wasn’t a chatbot. It was a workflow engine with opinions, apparently.

Where the demo shined... and where it wobbled

The strengths were clear:

Concise prioritisation: The agent didn’t ramble. It delivered a tight list of what mattered today.

Cross‑workspace synthesis: It pulled from Gmail and Drive like a seasoned project manager.

Project tracking: For the sprawling RTX Spark campaign, it navigated workstreams and lanes with surprising confidence.

But then came the wobble, the kind that instantly reminds you this is still early‑stage tech. Mid‑brief, the agent stalled. Just… stopped. With the star of the show stuck in a doom loop, the presenter shrugged, reran the request, and moved on.

It wasn’t catastrophic, but it was telling. Especially when paired with the admission that the presenter doesn’t use the agent day‑to‑day because their work laptop doesn’t have an RTX GPU (or indeed RTX Spark, which still hasn't been released commercially... or even given an estimated price range).

There has been no formal reliability testing, and there are still no firm benchmarks. So this was very much a capability showcase, not a validated product.

In other words: impressive demo, unproven reality.

Can you trust AI with your workday?

I found myself compelled to raise my hand and ask:

How reliable is this thing? Does it actually do the work, or does it take shortcuts to finish the task?

It’s a fair concern. Agents are notorious for skipping steps, hallucinating summaries, or confidently presenting half‑truths as gospel. And when your “chief of staff” is pulling from your inbox and your decks, accuracy isn’t optional; it’s the whole job.

The presenter’s answer? Essentially: We don’t know yet.

That’s not a deal‑breaker in itself. This is early tech, after all, but it’s the difference between a cool demo and something you’d trust with your caffeine-deprived Monday morning.

Running locally on an RTX GPU is a genuinely compelling angle:

Privacy: Your emails and documents never leave the device.

Latency: No cloud round‑trips.

Offline capability: Your chief of staff works even when your Wi‑Fi doesn’t.

Customisation: You can tune the model to your preferences.

But local also means:

Hardware dependency: No RTX hardware, no agent.

Maintenance overhead: Updates, model management, security: all local. With the hardware, software and electricity costs that entails.

Enterprise headaches: IT departments love centralisation; this is the opposite.

There's a definite trade‑off: privacy and speed versus accessibility and scalability.

If NVIDIA can deliver reliability, this kind of agent could be transformative:

Executive prep is easier: Instant briefs, deck pulls, and meeting context.

Project orchestration is smoother: Automated chasing, summarising, and nudging.

Inbox triage is handled for you: Prioritisation based on actual relevance, not timestamps.

Real calendar intelligence: Conflict resolution, prep summaries, and proactive planning.

This isn’t about replacing humans. It’s about removing the glue‑work, the tedious admin that eats half the workday.

But that future depends on some important elements: Transparent source citations, error recovery when tasks stall, guardrails against shortcut‑taking, benchmarks that prove reliability under load, and user studies that show real productivity gains.

And as of right now, those pieces aren’t in place.

A promising vision with homework left to do

The demo was compelling — not because it was flawless, but because it showed a direction that actually makes sense. A local AI agent that understands your workday, respects your data, and handles the boring parts? That’s a future worth rooting for.

But the gaps were impossible to ignore: No reliability data, no accuracy benchmarks, no real‑world usage, and a mid‑demo stall that felt a little too on‑the‑nose.

This isn’t a chief of staff yet. It’s an intern with immense potential; but it's one who still occasionally freezes mid‑task and needs you to repeat the instructions.

If NVIDIA can close the reliability gap, this could become one of the most practical uses of local AI we’ve seen. Until then, it’s a promising prototype with a lot of homework left.

Attempting to prove why "it's over for Hollywood," one confident X user shared an AI-generated short emulating New York City in 1993, and while many praised the 'authentic' look, others soon noticed the hallmarks of AI slop. The technology may be getting more refined, but it seems we've got a ways to go before we can shake the notion that AI filmmaking is a gimmick.

"If this is the level of AI creations, then it's over for Hollywood and big production houses," claimed X user @john_my07, sharing an AI-generated video of early 90s NYC emulating the look of an HD camera. With warm, hazy colour grading and ambient city humdrum, it almost feels authentic, but there's an undeniable roboticness that gives it that soulless AI feel.

The usual suspects are present, of course. Extra fingers and gibberish signage are key markers of AI slop. Pedestrians walk in an orderly hive, doppelgängers walk side by side, and bodies merge straight into moving traffic, but the worst offence has to be the dual Empire State Buildings. "Not the Empire State Trade Centre," one commenter responded. "Every single one of these videos is always like 'look what AI can do!' And it’s just derivative of a movie a human being made," another scathingly added.

While it's easy to get swept up in AI existentialism when we see the technology evolving so rapidly with seemingly no way to counter it, it's worth remembering that a large majority still crave analogue creativity. With AI creating a divide among creatives across the industry, design instinct and experience matter more than ever. Despite the torrents of slop, human creativity is still invaluable and vital to the creative sphere.

These true wireless stereo (TWS) buds deliver a well-balanced tuning that offers a thick-sounding lower end, keeps things relatively neutral in the midrange, and adds a touch of spice up top. It offers a far more balanced approach to tuning than 90% of the TWS out there, and at the new price, I think this is one of the better deals in audio right now.

Of course, if you want to explore more over-ear wireless headphones, I've got you covered. But if it's really earphones you're after, I highly recommend the Status Pro X.

Status Audio's Pro X deep dive

At the heart of the Status Audio Pro X is a triple-driver configuration, combining a 12mm dynamic driver for deep sub-bass with dual Knowles balanced armatures for highs and mids. the resulting tuning is best described as balanced, with a weighty sub bass being partnered with splashes of high end, but with a rich midrange presence making this an engaging V-shape sound signature.

When put head-to-head with mainstream heavyweights like the Sony WF-1000XM5, the Pro X provides a noticeably more organised sound profile with impressive instrument separation – things are less stuffy and the bass bleeds less into the lower mids.

So it's perfect, right?! Well, no, and while the hybrid active noise cancellation reaches up to 52dB and handles steady low-frequency background drone well, it still trails the market-leading ANC found on the Apple AirPods Pro 2 (currently down to $199), especially when dealing with high-wind noise.

The companion smartphone app, while offering a customisable 8-band EQ, also feels slightly less refined than Sony's polished ecosystem. Nevertheless, for music lovers prioritising the best sound possible, I think the Pro X remains a standout choice.

Somehow, it's September again. And in the tech world, at least, that can only mean one thing: new iPhones are coming. iPhone launches have tended to feel a little incremental of late, notable upgrades giving way to minor spec increases. But that could all be about to change at Apple's 'Surprise and Shine' event that's happening tomorrow.

For one thing, Apple has a brand new CEO. John Ternus has officially replaced Tim Cook at the helm, and will no doubt be keen for his first major Apple launch to make a splash. We've already heard tell that Ternus is planning to shake the brand up from a design perspective.

And it already sounds like we're in for some revolutionary new products this year. Likely to steal the show is the the iPhone Ultra, Apple's long rumoured folding iPhone. Meanwhile, the iPhone 18 Pro is set to feature an impressive new camera array. And there could be some wildcards, such as AirPods with... cameras?

Keep this page refreshed for every announcement as it happens.

How to watch

You'll be able to watch the event right here at 10am PT / 6pm BST at the link below. And until then, we'll be bringing you all of the last minute rumours as they come in.

You can watch the livestream above on our site, or on Apple's YouTube channel. Alternatively, you can watch it on Apple's website or on the Apple app.

Apple Surprise and Shine: The main expected announcements

It isn't just iPhones that could be announced tomorrow. The iPhone event usually comes with a cohort of other tech launches like AirPods or the Apple Watch. You can see a full list below:

• iPhone 18 Pro & iPhone 18 Pro Max: The chatter around Apple's next Pro-tier handsets centres on a jump to a 2nm A20 Pro processor, a shrunken Dynamic Island notch, roomier batteries, and a variable-aperture camera that's tipped to stay exclusive to the larger Max model.
• iPhone Ultra: Apple's foldable debut is finally said to be nearing reality – think a slender book-fold hinge, a near-8-inch internal panel, a return to Touch ID over face-scanning, and a cost that reflects its status as a first-generation device.
• Apple Watch Series 12: The everyday Watch line is rumoured for a refresh built around new health-monitoring tools, a shake-up of screen dimensions, and quietly upgraded internals.
• Apple Watch Ultra 4: Leaks point to Apple sharpening its adventure watch further still, with stronger satellite coverage, a battery built to go the distance, and more capable onboard sensors.
• AirPods 5: Skip the camera-equipped AirPods Ultra for now – that's for 2027. This one's a lighter-touch update to the everyday earbuds, sticking close to the AirPods 4 shape while pushing Siri's AI power as the main selling point.
• HomePad: Apple's rumoured stab at a smart-home control panel reportedly combines an iPad mini-sized screen, a built-in FaceTime camera, and sensors that clock who's in the room to trigger automations.
• HomePod mini 2: This pint-sized speaker has been waiting a while for its turn – expect a sound quality bump and snappier on-device Siri handling in the same footprint.

The M56C treads a very fine line for Sihoo: it's one of its cheapest chairs ($180 / £140), and it trims back the extras as much as possible to keep the price low. You won't find an adjustable lumbar support or flip-up armrests here, the assembly process is convoluted, and there are times when it reminds you just how cheap it is.

While the M56C won't be gaining a place in our round-up of the best office chairs for back pain, it's an intriguing option that could find a home in certain situations.

Comfort & ergonomics

After many years of examining office chairs, I’ve found that they fall into roughly two camps: they’re either springy and energetic or plush and laid-back. Usually, cheaper chairs come in at the bouncier end, rebounding to your movements like a trampoline, but the M56C is surprisingly plush: you sink into its memory-foam cushion as you sit on it, and the supportive mesh has more give than other chairs. This is no bad thing, and it makes it feel like it’s much more expensive than its price tag and build might suggest.

Sihoo has pared back the adjustability to the bare minimum, though, adopting a universal approach that means anyone can sit on it — that is, if you’re 5’ 2” to 6’ 2”. At 6’ 4”, I’m just about out of Sihoo’s recommendations, but at its maximum extension and angle, the headrest aligned with the top of my neck (as it should), and the armrests never felt too high for my noodly limbs.

Adjustability is not a priority — the M56C keeps things cheap and universal, making it a good chair to share. The headrest is the most movable thing here, clicking into position vertically and angling through 45 degrees. The armrests only move up and down, but they do lock into place securely. The backrest reclines and locks from 110 to 125 degrees, which isn’t a huge amount, but enough to make it feel slightly more comfortable.

The chair is comfortable in places, with the seat pad being particularly cushioned and enveloping, and the non-adjustable sprung lumbar support proving just right. However, there are some areas that feel distinctly unrefined. While most armrests are made of foam, the M56C’s seem to be hollow plastic with very little give, which can be hell on bony elbows. And if you lean far back into the chair and the mesh hits the plastic support, digging into your back uncomfortably.

The initial plushness of the M56C declines as the day goes on. It’s a chair that's suited to short bursts of work or quick meetings over day-long sessions.

Design & aesthetics

Do you like your chairs to be black? Well, you’re in luck! The M56C is available in black, and black only. It is the most common colour for office chairs; it doesn’t stain and it slots in effortlessly, but I do like to see a range of at least one other colourway. I may be a little biased here because it’s hard to photograph black chairs without them looking like amorphous blobs.

Colour (or lack thereof) aside, the M56C is a conventional chair that doesn’t look too chunky or squat like some more ergonomic chairs. It lacks the structured grace of Sihoo’s M18, but the mesh fabric always has a certain airiness that doesn’t look quite as oppressive as alternatives. Even in black.

Specs

Assembly

Hoo-wee! The assembly process for Sihoo chairs is usually fairly straightforward, but the M56C proved to be surprisingly complicated. There are a lot more parts here, which presumably keeps the whole thing cheaper and more compact when boxed up, but it also adds a few more steps to the build.

It could be simple to put together if it weren’t for the dodgy manual and the bolts being supplied in plastic bags rather than a labelled blister pack. There are mislabelled steps, very similar-looking bolts, and a reference to a component that looks completely different in real life. I guessed my way through the assembly process and it felt like I arrived at the finished chair through coincidence rather than following instructions.

Price & value

The M56C is priced at around $180, though I've seen it sell for a few dollars cheaper. The closest comparison here is Sihoo’s own M18 ($159), which has become our absolute favourite budget office chair. They’re (unsurprisingly) similar chairs, but the M18 gets so much right that the M56C doesn’t quite land: straightforward assembly, a one-piece backrest without hard edges, and an adjustable lumbar support that’s actually one of the industry’s best.

Sihoo’s M59AS ($239) is also similar, with an exciting springiness and flip-up armrests for active people who frequently switch between tasks.

Moving up the scale, Boulies EP200 ($279.99) and Branch Ergonomic Chair ($339) are pricier, but add style and more adjustments, so you can feel supported for longer sessions.

Should you buy?

The M56C is a misstep from the usually reliable Sihoo, leaning too far into low-budget construction and missing the mark in terms of comfort, with hard armrests and stiff plastic supports making it uncomfortable for longer sessions.

It’s still a good-looking chair, and its one-size-fits-all approach, comfortable cushion and low price make it a good choice for kitting out a meeting room where people don’t have to sit for too long. For everywhere else, Sihoo’s M18 still rules the roost.

The presenter — Rick, armed with two identical Surface laptops and a healthy dose of confidence — showed Blender’s Cycles viewport running DLSS 4.5 side‑by‑side with OIDN and OptiX. The claim was simple: DLSS 4.5 delivers “instant feedback,” while the traditional denoisers make you wait for the image to resolve.

If you’ve ever nudged a light in Cycles and watched your scene dissolve into a grainy snowstorm before slowly reassembling itself, you know exactly why this matters.

But as with all things AI‑accelerated, the promise comes with caveats — and creatives are right to keep one eyebrow raised.

The demo: two laptops, one message: speed wins

Rick’s setup was straightforward: we had Blender’s default OIDN on one laptop, with OptiX on the same hardware, while we got DLSS 4.5 with ray reconstruction on the other.

OIDN is the reliable workhorse. OptiX is the NVIDIA‑optimised speedster. But DLSS 4.5? It was pitched as a different beast entirely: not just denoising, but reconstructing ray‑traced detail using a learned model, complete with super resolution and a mysterious “transformer version 2” under the hood.

The viewport looked smoother. More stable. Less like a flickering cloud of noise pretending to be your scene. And crucially: changes appeared (nearly) immediately.

For look‑dev artists, lighting TDs, and anyone who spends their day nudging sliders and rotating HDRIs, that responsiveness is gold. Creative flow thrives on immediacy. Every second spent waiting for samples to accumulate is a second where your idea cools off.

Why this matters: creative flow is not a luxury

We talk a lot about render times, but not enough about iteration times. The viewport is where ideas form — where you experiment, break things, fix them, and try again. When that loop is slow, creativity suffers.

DLSS 4.5’s pitch is essentially: What if Cycles felt less like a path tracer and more like a real‑time engine while you’re working?

That’s a big deal, not because it replaces artistry, but because it removes friction. And friction is the enemy of bold creative decisions.

If DLSS 4.5 genuinely delivers smoother previews without compromising the accuracy artists rely on, it could meaningfully change how fast teams iterate, review, and approve work. But — and this is where I have to put my editorial hat on — speed is not the only metric that matters.

Faster isn't always better

Artists have learned to be wary of AI‑powered “enhancements,” because they often come with trade‑offs that only show up once you’re knee‑deep in a production scene.

Some open questions:

Does DLSS 4.5 stay faithful enough for lighting and shading decisions?

Instant feedback is great, but not if it lies to you.

How does it handle thin geometry, fine textures, or low‑sample chaos?

Denoisers love to hallucinate detail. Reconstruction models sometimes hallucinate too much.

Will it behave consistently across stylised, NPR, or heavily composited scenes?

Blender artists don’t all work in photorealism.

Is this NVIDIA‑only acceleration going to widen the hardware divide?

OIDN works everywhere. DLSS does not.

And then there’s the bigger industry concern:

Does faster iteration become an excuse for tighter deadlines and smaller teams?

Technology that empowers creatives can also be weaponised against them. We’ve seen it before.

The tech context: DLSS isn't just a denoiser any more

DLSS 4.5 with ray reconstruction isn’t simply “OptiX but faster.” It’s a learned model that replaces parts of the denoising and temporal filtering pipeline entirely. In theory, that means better stability, cleaner glossy reflections, fewer temporal artefacts, and more detail retention in low‑sample previews

If Blender 5.3 integrates this cleanly, it could become the default viewport experience for RTX users. Not because it’s flashy, but because it’s useful.

But usefulness depends on trust. And trust requires independent benchmarks, not a controlled demo with unknown scene complexity.

Great for RTX users, [shrug emoji] for everyone else

Let’s be honest: this is an NVIDIA‑only party.

OIDN remains the cross‑vendor baseline. OptiX is NVIDIA‑optimised but still broadly accessible. DLSS 4.5, however, is locked to RTX hardware.

If Blender’s viewport experience becomes significantly better on NVIDIA GPUs, that’s great for RTX users, but it risks creating a two‑tier creative ecosystem unless AMD and open‑source alternatives catch up with their own reconstruction models.

Blender has always been a champion of accessibility. DLSS integration will need to be handled carefully to avoid fragmenting the user base.

DLSS 4.5’s ray reconstruction looks like one of the rare AI‑accelerated features that actually respects the creative process. It doesn’t try to automate artistry. It doesn’t replace human judgement. It just makes the viewport less annoying.

That’s a win.

But creatives are right to be cautious. Faster previews don’t automatically mean better decisions. And AI‑powered reconstruction models can be unpredictable in ways that only show up after hours of work.

The promise is real, and the potential is exciting. But like with AI video enhancements, the industry must resist the temptation to treat speed as a substitute for craft, or as a justification for squeezing artists harder.

If DLSS 4.5 lands cleanly in Blender 5.3, it could become one of the most meaningful quality‑of‑life upgrades Cycles has seen in years. Just don’t let the marketing gloss distract from the need for rigorous testing, cross‑vendor parity, and respect for the creative labour this tech is supposed to support.

The key element of this artwork is the boss design, and this is where I started. I established rough guidelines in the form of a concept sheet, using a lot of references for each design element I wanted to incorporate. With the design sorted out before the illustration, I was able to focus on storytelling through mood, colours and composition to make this showdown as epic as possible.

(Note that you might want one of the best laptops for 3D modelling or the best drawing tablets to work on).

Today's best drawing tablet deals

This article originally appeared in ImagineFX. Subscribe to ImagineFX to never miss an issue. Print and digital subscriptions are available.

Creative Bloq is now easier to access than ever before with our on-the-go app, which brings you all the content you know and love from our website, but in a super-streamlined design.
Download the Creative Bloq app for iOS
Download the Creative Bloq app for Android

But a less-often discussed impact is what AI's doing at the very beginning of the process, when a business is looking for an agency to develop its brand identity. The founders of small startups used to write personal briefs that could win them a talented creative agency at reduced rate if the project was interesting enough. AI-written briefs may be ending that.

Small design agencies have long been quietly subsidising the startup ecosystem. If a founder with big ambitions but limited budget pitched an interesting idea, an agency might take a risk on them if they liked the concept, hoping that relationship would pay off further down the line or provide a good portfolio project in a new area.

Daniel Hinde, founder of London agency Greatergood® Brands , argues that this approach has seen billions of first-time brands along the road to launch, but the it's now breaking.

"The nature of the briefs we're getting from startups has changed completely in the last year," he says. "It's gone from 'we're a startup and we only have X amount of money' to briefs that were clearly never written by the founder at all. They've been generated by an AI in a matter of seconds, and bulk sent to every agency sitting at the top of a Google search."

It's easy to see why founders would be tempted to use AI. Creating a brief to approach an agency requires time and thought. A chatbot can do it in seconds. You give it information about your product and serve, and it will generate a brief based on what it believes such a document should look like according to its training.

The problem is that there's no standarised brief that fits every business. And the kinds of things that big corporations need are very different to what a one-person startup needs to launch their brand.

"They're completely over-egged," Daniel says of the AI-written briefs he's received. "They'll ask for a full strategic proposal covering brand architecture, brand design, brand toolkit, packaging, retail strategy, go-to-market launch strategy, a website, a content plan, for a company that is genuinely one person and a £14.99 subscription to an AI model.

"A founder this size doesn't need a fraction of that yet, and often can't afford even one element of it, but the AI doesn't know that. It just writes the most comprehensive brief it can generate and fires it out."

AI can write the brief in seconds, and the startup can sent it out to dozens of agencies simultaneously. That might feel like efficiency, but, shooting into the void like that is unlikely to get the attention of the right agency. By using AI, business founders are forgetting that many of the best creatives are discerning when it comes to choosing clients: you need to sell yourself to them as much as the other way around.

"When a brief used to land from a scrappy founder, you could tell someone had actually sat and thought about their business," Daniel says. "That's exactly the kind of project we'd take a chance on. Most agencies will take a punt on a startup, perhaps the piece of portfolio work would be good and open a new category or new doors, perhaps we even just liked the founder on a personal level.

"Now nearly 90% of what lands from startups is impersonal, identical to what every other agency in the top ten search results is receiving, and asking for ten times the scope that startup actually needs. It's much harder to tell which of those founders is worth betting on."

“In addition, the AI models all have their set of rules for how we agencies should respond to this life-changing opportunity heading our way. At the end of each “brief” is the process and list of everything required to send our proposal. The result is that serious agencies will simply send a quote at the same costing which we would send to a global client group, not a startup. It’s either that, or we don’t respond at all.” – says Hinde.

It seems this is another area where a technology intended to make things faster and easier could be making things harder for the people it was supposed to help. Whether that trend reverses depends on business founders themselves. Daniel suggests the fastest way for a founder to make an impact on a creative agency in 2026 might be to simply write their own brief.

Daniel Hinde is founder and creative director at Greatergood Brands – a brand and packaging design agency who launch scale, and reposition consumer brands through strategic brand and packaging design.

Despite the super-human powers given to him by the Greek god of the same name, and a wild costume even for the golden age of comics, Atlas, AKA Jim Randall, failed to reach the lofty heights attained by the likes of Hawkeye and Daredevil, which emerged in the same year.

But one creator is giving him the second chance he (maybe) deserved – and he's put a lot into it. The filmmaker and YouTuber Austin McConnell has spent four years working to bring Atlas to life in a 90-minute 3D animated movie. That involved having to learn pretty much the entire animation pipeline, Austin explains in the video below.

Austin was looking for lesser-known golden-age superheroes that had entered the public domain to create a low-budget live-action "Superzeroes" cinematic universe as an alternative to the many polished superhero franchises. Atlas grabbed his attention because of his extreme obscurity and his ridiculous costume.

Austin decided that, relatively speaking, animation would be a more practical medium to tell the character's story. 'Relatively', because at the time he had only had a basic knowledge of animation from adding quick CG effects for YouTube videos.

“I had to learn the ins and outs of 3D animation, clothing simulation, hair simulation, motion tracking,” he says in an interview with Cartoon Brew. “I bought a mocap suit and had to learn how to set all that up. I had to do real-time facial tracking.”

Despite the steep learning curve, Austin went on to do most of the animation himself, working six or seven hours a day in his basement and hiring people for specific tasks as needed. He mainly used Reallusion’s iClone (see our guide to the best animation software), with After Effects, Photoshop and DaVinci Resolve for editing, effects and creating art assets.

A Kickstarter campaign raised $45,100, most of which was used to hire voice actors and buying gear. The CG artist Kalex3D helped redesign Atlas's costume and model characters and later stepped in to animate some parts of the film after Austin suffered a stroke in 2024.

His limited experience in animation led Austin to lean into other aspects, like pacing editing. At one point, he had to make a decision to stop aiming for perfection and embrace the "original spirit" of the project as a low-budget movie with heart. Workarounds included having a character drink off camera to avoid having to animate the movement.

"It was really important for me, knowing my limitations in terms of animation skills, that the other parts supplemented that,” he told Cartoon Brew. “Mine is kind of the inverse of the traditional indie animation problem. I bring a better sensibility in terms of editing, audio production quality, and pacing, and the animation is sort of tacked on afterward.”

Austin found that the rough, choppiness of his approach worked for the movie's narrative. He was determined not to use AI in order to keep a homegrown feel, and the movie benefits from that.

There's been some debate about whether opposition to AI is the reason the bizarre Chinese blockbuster Niu Lai has done so well. Atlas isn't as raw looking as that – and it's actually a good movie – but its less polished animation suits its character. After all, Atlas was a bargain basement comic book hero rescued from a defunct publisher, not a glossy, Marvel character like Daredevil, or even Deadpool.

"No one is going to think this is AI-generated because there are scenes where it’s like, ‘Yeah, this was one dude making this on his home computer’," Austin says. "That’s the spirit in which it was made, and we try to play that to our advantage.”

He's already developed more stories in the same universe, including a second animated film slated for later this year.

“If people want to see more from these characters and these stories, I’m happy to make more. If nobody is really that interested, I’m happy to move on to something else,” he says.

Atlas: The Animated Movie goes live on Austin's YouTube channel today (8 September) There will be a screening at Moxie Cinema in Springfield, Missouri, on 12 September.

Inspired? See our guide to the best laptops for animation.

You've heard of 'iPhone pinky'; whereby one's little finger becomes sore as a result of resting a phone on it during a session of doomscrolling. But according to one physical therapist, the extra weight of the so-called iPhone Ultra could be bad news for our thumbs.

Leaks suggest the foldable iPhone could weigh up to 78g more than the standard iPhone 17 when folded, with the extra weight ramping up strain on users' hands. Musculoskeletal physical therapist Martin Ratcliffe recently told mobile phone recycling comparison site Compare and Recycle about what this means for your hands:

“After several hours of use, the continued strain from the added weight can lead to fatigue and increase tension on the tissues, particularly the tissue around the thumb, which is the main finger used to navigate devices. Furthermore, the grip position may change because of the phone's foldable design. Overall, although the additional weight may be small, the repetitive strain and low-grade tension may lead to further stress and injury to surrounding tissues and joints.”

This all might sound a little alarmist, but there's precedent for foldable phone users complaining about hand strain. On a Reddit page for Samsung Galaxy Fold owners, users have asked whether others have experiencing aching or stinging hands.

Martin says to take breaks every 30-45 minutes to reduce the risk of tendon and muscle strains: “During these breaks, I recommend people put the phone down and look away to ensure a proper rest. People should stretch their palms and shake their hands for a few seconds to promote increased blood flow. Stretching each digit supports muscle repair and reduces tension.”

That's the pro choice for anyone ready to make the move to a larger tablet with ultra-low latency, but there are plenty of deals about for beginners too, including on some of the devices we recommend in our expert guide to the best drawing tablets.

These are our picks.

Frame TVs are two products in one. You get the features you'd expect from a modern smart TV for streaming movies and series, with the benefit of being able to display wall art on the canvas-like display when you're not watching TV. Most let you display your own art or offer a subscription to a library of art that you can bring into your living room and rotate as desired.,

If you have a large space to fill, there are some great savings on larger frame TVs from highly rated rivals, including the Hisense Canvas, which we rated as the best Frame TV overall when taking price into consideration, and the LG Gallery. And if you're interested in displaying digital art but not in the TV part of Frame TVs, there are also a few great Labor Day deals on some of the best digital frames.

Labor Day digital art frame deals

Beiluner is one of the most popular pen brands on Amazon for a reason. Its luxury fountain pens feature a classically elegant design, with a 24k gold finish and German Schneider ink absorber. The result is a writing experience that looks and feels more luxury than it costs.

And right now it costs even less. For a limited time, the Beiluner luxury fountain pen is 47% off at Amazon, bringing the price down from $37.99 to $19.98. Reviews rave about the luxury feel and value for money – which, at under $20, is better than ever right now.

But if floor-shaking bass isn't your priority, there are plenty of great steals in Amazon's Labour Day sale. From sleek portable options to powerhouse rigs built for partying, there's a range of styles to suit your sound spec needs and even your interior design aesthetics.

Can't spot anything that takes your fancy? Check out these speaker deals sorted by region.

Not because they’re magical, but because they’re practical.

An AI feature that doesn't break the creative flow

Most AI tools promise speed, but they often demand a whole new workflow, a new app, or a new mental model. Creatives don’t want that. Creatives want to stay in Photoshop, stay in their groove, and stop wrestling with the technical busywork that interrupts the fun part.

The demo I saw at NVIDIA's RTX Spark showcase at IFA in Berlin showed a sketch of a robot turned photoreal using a beach photo as a main reference, with smaller reference photos and images to inform the model on all the individual elements of the robot, including the colour, texture of the metal, shape and type of 'eyes' and more. It wasn’t impressive because it was flashy. It was impressive because it respected the artist’s process. In fact, the demo was presented by the artist herself. The agent didn’t overwrite the sketch, reinterpret the brief, or decide the robot should be a cat now. It simply did the heavy lifting the artist didn’t want to do.

That’s the first time in a while I’ve seen AI behave like a collaborator at an advanced level, instead of a chaos gremlin.

Local AI is a feature creatives have been begging for

Let’s be honest: the creative community has trust issues with cloud AI. And for good reason. Too many models have been trained on artists’ work without permission, and too many “helpful” features quietly upload your assets to servers you’ve never heard of.

This demo didn’t do that. Everything ran locally on a (rather beefy, admittedly) RTX workstation with 128GB of unified memory. No cloud. No mystery training. No “your PSD is now part of a dataset in Utah”.

The presenter said it plainly during the demo: “My creative data stays on my machine.”

That alone makes this one of the most creator‑friendly AI features we’ve seen in years.

Iteration at (nearly) the speed of imagination

The real magic wasn’t the photoreal robot. Rather it was how quickly the artist could riff on it.

“Change the theme to nightlife.” Boom: neon nightlife in ten seconds.
“New camera angle.” Hey presto: a generated 3D model you can spin around.
“Make a video between these angles.” In under a minute: a workflow that used to require three different apps.

Sure, one version had cars floating in the ocean. Generative AI still loves a bit of surrealism. But when the entire workflow takes 90 seconds, you can afford to laugh, delete, and try again.

This is the kind of speed that encourages experimentation instead of punishing it.

The real reason this matters, is that these AI agents inside Photoshop aren’t about replacing artists. They’re introduced in an aim to reduce friction. They’re about letting creatives stay in the zone longer, about giving you a superpowered assistant who doesn’t get tired, doesn’t misplace files, and (hopefully) doesn’t ask what the brief was again.

If Adobe plays this right - and importantly, keeps the local‑first approach - AI agents could become the most meaningful workflow upgrade Photoshop has had in a long time.

Not because they’re clever, but because they finally understand what creatives actually need.

I’ve been digging through the deals and pulled together some of the more useful upgrades, from big desk mats and monitor stands to mechanical keyboards, desk lamps and charging docks, along with a few things that are simply nice to have when you spend a lot of time at your desk. There are options for artists and designers too, including some sketching gear, so you don’t necessarily need to build an elaborate dream setup to find something worth buying.

That’s the appeal of these deals: you can make your desk feel considerably better without having to replace the expensive stuff underneath it, though these seven Labor Day laptop deals are worth checking out.

The pitch is deceptively simple: generate half the frames, let RTX Spark-powered local AI infer the rest, and walk away with a 60 FPS clip that looks like you spent twice the compute. It’s the kind of optimisation that makes engineers grin and creatives squint suspiciously.

Because yes, the tech is clever. And yes, the speed gains are real. But the implications for creative workflows - and creative labour - are more complicated than a neat demo on a show floor can fully address.

Let’s dig in.

The promise: generate less, infer more, but without the usual AI weirdness

The core idea behind RTX Spark’s pipeline is almost cheeky in its simplicity: Generate at 30 FPS (and at a lower resolution) using NVFP4 precision on a Blackwell-architecture unified memory chip (the RTX Spark platform). Then hand the output to an RTX video node that handles AI upscaling and AI frame generation. Finally, deliver a 60 FPS, higher‑resolution final video that looks like you did twice the work.

If you’ve ever used DLSS in games, you already understand the logic. But this isn’t real‑time interpolation; it’s handled entirely offline (using the RTX Spark tech powering NVIDIA's new unified memory), with more time per frame and access to full motion flow. That means cleaner in‑betweens, fewer temporal glitches, and none of the plasticky shimmer that makes some AI‑assisted footage look like it’s been dipped in varnish.

In other words: the pipeline doesn’t just save time. It saves face.

NVFP4 + Blackwell: an efficiency play that creators will welcome?

NVFP4 is the unsung hero here. It’s a lower‑precision format that Blackwell chews through like popcorn, giving you noticeably faster inference, lower compute cost per frame, and little or no visible quality loss for typical content.

This matters because AI video generation is still compute‑hungry enough to make even well‑funded studios wince. If you’re an indie creator, a small production house, or anyone who doesn’t have a server farm in the basement, shaving minutes off every iteration isn’t just convenient; it’s the difference between “we can try 3 versions” and “we can try 30”. It promises real workflow improvements.

The numbers: a fifth of the render time

The benchmark from the IFA demo was eye-opening indeed:

- Full HD, two‑second clip: about 6 minutes 30 seconds

- Same clip with the half‑frame pipeline: about 1 minute 25 seconds

Even generating every frame in NVFP4 would have cut the render roughly in half. But the half‑frame strategy pushes the savings into “this changes how we work” territory.

For short‑form creators, the TikTok/YouTube/marketing crowd, this is huge. For animation teams iterating on motion, lighting, or composition, it’s possibly even bigger. And for anyone experimenting with AI video as part of a hybrid workflow, it’s the difference between “interesting toy” and “practical tool”.

The creative anxiety: speed is great, sure. But what about quality?

Where it holds up:

Weinand claimed no visible difference in the end result between a fully generated 60 FPS clip and a 30 FPS clip upscaled and frame‑generated to 60 FPS

And in the demo, that seemed true. Motion was smooth, textures held up, and the dreaded “AI slop gloss” didn’t rear its head.

Offline interpolation is simply better positioned to avoid the uncanny valley than real‑time frame generation.

Where creatives should still be cautious:

This isn’t a magic wand, and creatives know better than to trust a demo blindly. Low base FPS (10–15) upscaled to 60 will absolutely produce artefacts. Fast‑motion content, such as sports, dance, or chaotic handheld shots, may need higher base FPS. Fine textures (hair, fabric, detailed animation) can still trip up interpolation. And stylised content may behave unpredictably depending on how the model interprets motion flow.

And then there’s the bigger concern: consistency. AI pipelines are notorious for behaving beautifully in one scene and bizarrely in the next. Creatives aren’t worried about the best‑case scenario; they’re worried about the worst‑case one that ships.

Does this matter more than any other "AI breakthrough"?

RTX Spark isn’t just a speed trick. It’s a sign of where AI video is heading: toward hardware‑accelerated inference, modular pipelines, and hybrid workflows where generative models don’t have to do everything themselves.

This is good news for creators who want AI to be a tool rather than a takeover.

But it also raises the stakes. Faster generation means faster iteration, which means faster production cycles, and that inevitably feeds into the ongoing debate about creative labour, job security, and the role of human artists in an increasingly automated pipeline.

The tech is promising, but the implications are messy.

The ecosystem is already here: RTX video capabilities already exist in consumer tools like VLC, professional software across video and VFX, gaming pipelines, and CUDA‑based pro workflows.

The GitHub release rolling out post‑IFA includes AI super resolution (available now), AI frame generation (coming shortly) and an HDR module (in development).

RTX Spark’s half‑frame pipeline is one of the first AI video features that feels genuinely creator‑friendly. It doesn’t try to replace artistry. It doesn’t bulldoze workflows. It doesn’t promise magic. It just makes the slow parts faster.

But creatives are right to be wary. Speed is seductive, and the industry has a habit of using “efficiency” as a cudgel. Tools like this can empower creators, but they can also be used to justify unrealistic deadlines and smaller teams.

The tech is impressive, and the potential is real. The responsibility lies with how the industry chooses to use it.

But what's next?

How does this pipeline behave with sports footage, high‑detail animation, or VFX‑heavy sequences? What presets will emerge as the “sweet spot” for balancing FPS, resolution, and upscaling? How will the HDR module fit into professional colour workflows?

And most crucially: how will studios integrate this without destroying the human side of the creative process?

AI video is evolving fast. RTX Spark’s approach feels like a step toward a future where creators can work smarter without sacrificing quality... but only if the industry remembers that speed is a tool, not a mandate.

While the design world is no stranger to logo disputes, Maryland University's fierce cease-and-desist strategy proves just how important heritage design can be. With the dispensary not willing to part with its identity, it seems we're entering the ultimate battle of the script fonts.

While the university has historically used the term 'Terps' as shorthand for its sports teams, the Maryland Terrapins, the word has taken on a very different meaning in the cannabis industry. The rival Terps USA brand name comes from terpenes used to modify and enhance cannabis products, and while the name clash seems like a coincidence, Maryland University isn't pleased with the branding.

Both the Terps and Terps USA branding feature a cursive font in bold red, with allegedly similar flourishes on the 'r' and 's' letterforms. In an email to Terps USA owner Ken Fry, the University claimed, "Your company has stolen the name and trademarked image from the University of Maryland," with the cease-and-desist letter soon to follow. "Such an association would materially impair the goodwill embodied in the University’s famous TERPS marks,” it claimed.

Ken has since adjusted the logo, keeping the cursive font with the addition of a gold drop shadow embellishment. The red has been swapped for a more fitting green theme, but the owner is keen to maintain the name, claiming, "It’s ours. You’re not taking our name because you don’t like it." Ken went on to tell Fox 5 DC, "Will we sit down and talk to you about the logo and have a conversation and work it out? Absolutely. We’d be more than happy to do that. But we’re not giving up our name."

For more design disputes, check out the Patagonia logo lawsuit controversy that is every brand's worst nightmare, or take a look at when Aston Martin picked a logo battle with its own shareholder.

The best deal I'm seeing right now is $200 off the K2 Combo 3D printer model, which brings the price down to just $499. This machine is ideal for beginners or anyone dipping their toe into multicolour 3D printing, thanks to the included CFS for auto colour switching. For something more pro-level, there's a very nice $300 off the K2 Plus Combo – now $1,199 at Creality.

See our guide to the best 3D printers for some more recommendations, and don't forget to check out Creality's range of the best laser cutters as well as the best 3D scanners too.

• Shop the Creality Labor Day sale