If you work with images or video professionally — or even just obsessively — this week’s announcement from Topaz Labs is worth stopping everything to read. On April 28, 2026, the Dallas-based AI imaging company dropped what it’s calling its Next-Gen update: the largest single release of AI models in the company’s history. Six new models across both photo and video, a brand-new proprietary inference engine, and expanded web access — all at once.
Here at Creative AI Show, we’ve been tracking AI creative tools closely, and this one caught our attention immediately. Here’s everything you need to know about what just shipped, what it means for your workflow, and why it signals a major shift in how AI-powered image and video enhancement is being built.
What Is the Topaz Labs Next-Gen Update?
Topaz Labs has been a fixture in the professional photography and videography toolkit for years. Tools like Topaz Photo AI and Topaz Video AI have become go-to solutions for photographers who need to rescue underexposed, noisy, or low-resolution images, and for editors who need to upscale and enhance video footage without destroying detail.
The Next-Gen update is different from a typical incremental release. Rather than iterating one or two models, the company shipped four entirely new image AI models and two new video AI models simultaneously — with a new underlying technology stack to run them more efficiently than ever before.
The Four New Image Models
The heart of the Next-Gen update is a complete overhaul of Topaz’s image processing suite. Here’s what’s new:
- Wonder 3 — The flagship one-click model. Wonder 3 combines sharpening, upscaling, and denoising into a single pass, designed to produce professional-grade output with minimal configuration. If you want to hand your raw or slightly blurry image to an AI and get back something print-ready, this is the model for that. It’s also available in Topaz Express, the company’s web-based tool, which means you don’t even need the desktop app to access it.
- Denoise Max — A dedicated denoising model pushing the limits of what’s possible in noise and grain removal. Topaz describes this as the new standard for noise reduction, particularly for high-ISO shots and archival footage restoration. Like Wonder 3, it’s accessible through Express on the web.
- Super Focus 3 — The third-generation sharpening model, built for images with significant motion blur or out-of-focus areas. Super Focus 3 is designed to recover fine detail in difficult situations where traditional deconvolution sharpening would create artifacts or halos.
- High Fidelity 3 — The upscaling model aimed at preserving the maximum amount of authentic detail when scaling images up. This is the model you’d reach for when enlarging for large-format printing or billboard-scale output, where every pixel matters.
All four models are available in Topaz Photo desktop application and via the Topaz API, giving developers and enterprise users the ability to integrate them directly into custom pipelines.
Two New Video AI Models
The Next-Gen update isn’t just for photographers. Topaz Video AI — already widely used in post-production workflows, archival restoration, and broadcast — also gets two new models:
- Starlight Precise 2.5 — An evolution of Topaz’s Starlight model line, which is designed for enhancing low-light video footage. The 2.5 update brings improved precision in preserving grain character and fine texture even in very dark scenes — a common pain point when converting old film or night-shot video.
- Astra 2 — The updated general-purpose video enhancement model. Astra 2 is designed to handle a wide range of scenarios, from web video upscaling to archival footage restoration, with better temporal consistency (fewer flickering artifacts between frames) than its predecessor.
For editors working in DaVinci Resolve, Premiere Pro, or Final Cut Pro with Topaz Video AI as a plugin, these models drop straight into existing workflows without requiring any new integration work.
The Technology Behind It: Topaz NeuroStream
Six new models is impressive, but the most technically significant part of the Next-Gen announcement is the introduction of Topaz NeuroStream — a proprietary inference engine built to reduce VRAM consumption by up to 95% compared to running these models on standard frameworks.
That number deserves emphasis. A 95% reduction in VRAM usage means that models which previously required a high-end workstation GPU can now run on consumer-grade hardware — laptops with integrated graphics, older desktop cards, and mid-range systems that would have struggled with previous generations. For independent creators, photographers working from the road, and small studios that can’t justify $3,000+ GPU upgrades, this is a meaningful change.
NeuroStream also allows models to run locally on-device, which matters for users in privacy-sensitive workflows (medical imaging, client confidentiality in commercial photography, legal documentation) who don’t want to route images through cloud processing.
Topaz Express: Web-Based Access Gets More Powerful
Not everyone wants to install desktop software, and Topaz has clearly taken note. The company’s web-based tool, Topaz Express, now supports both Wonder 3 and Denoise Max — giving browser users access to two of the most capable new models without downloading anything.
For casual users or those just evaluating Topaz’s capabilities, Express lowers the barrier significantly. It also means team members who don’t have the desktop app installed can still process images in a pinch from any machine with a browser.
Why This Release Matters for Creative AI
The Next-Gen update isn’t just about better models. It reflects two broader trends in the creative AI space right now.
First: local AI is making a comeback. After years of everything moving to the cloud, there’s a clear market pull toward on-device processing — for privacy, for latency, and for offline capability. Topaz NeuroStream is a direct bet on this direction. If you can run powerful models on consumer hardware with 95% less VRAM overhead, cloud dependency becomes optional rather than required.
Second: specialization still wins. In a world where general-purpose AI video generators like Sora (now discontinued), Runway Gen-4, and Veo 3 dominate headlines, Topaz continues to carve out a different niche — AI that doesn’t generate new content, but makes existing content better. Enhancement over generation. For professionals whose work depends on authentic source material, that distinction matters enormously.
As CEO Eric Yang put it at the announcement: “This is both the largest and most technically advanced set of models we’ve released in the company’s history.” That’s a bold statement, but given what’s in the package, it’s hard to argue with.
How to Get Started
If you’re already a Topaz Photo AI or Topaz Video AI subscriber, the Next-Gen models are available now through the standard update. Wonder 3 and Denoise Max are also available via Topaz Express on the web.
If you’re new to Topaz, this is probably the best moment to try it — the combination of new models and the NeuroStream efficiency gains means you’ll get more capable results on whatever hardware you’re running.
And if you want to stay on top of releases like this as they happen, subscribe to the Creative AI Show on YouTube — we cover every major creative AI tool drop as it lands.
Bottom Line
Topaz Labs’ Next-Gen update is the kind of release that doesn’t always generate the loudest headlines — there’s no flashy demo video of AI generating photorealistic impossible scenes — but for working photographers, videographers, and post-production professionals, it’s directly relevant to daily work. Six new models, a 95% VRAM reduction, web access to the best new models, and API availability for developers — all in a single update. It’s a strong start to May 2026 for the creative AI space.
No responses yet