AI image enhancers are moving from “make it sharper” to full-blown image repair and cross-device pipelines. The clearest proof is that Topaz Labs, a longtime name in photo and video enhancement, rolled out what it called its largest single release of AI models in company history on April 28, 2026—adding four new image enhancement models and two new video enhancement models across multiple apps. At the same time, Vmake leaned into an entirely different use case with an Old Photo Restoration capability aimed at fixing faded, damaged, and low-quality photos, pushing a consumer-friendly restoration story ahead of Mother’s Day.
That combination—more model capacity plus more restoration-oriented workflows—points to a shift in how enhancement software is being built. It’s no longer just about optional polish; it’s about solving the hardest, messiest inputs: compressed, noisy, motion-ruined, color-shifted, and damaged photos. The best enhancers increasingly behave like “image recovery engines,” not filters.
Topaz Labs: Next-Gen models are about precision, not just punch
Topaz Labs’ “Next-Gen” launch on April 28, 2026 matters because it expands the model toolbox in a targeted way: the release includes four new image enhancement models with capabilities spanning sharpening and denoising, while also introducing two new video enhancement models. The practical implication for creators is that “enhance” can now be decomposed into more specialized stages rather than relying on a single end-to-end guess.
Why does that matter? Noise and blur are different kinds of failure. Denoising involves reconstructing lost detail from patterns in the noise distribution; sharpening, meanwhile, changes edge structure and micro-contrast. When a platform treats these as separate model jobs, it can better manage the classic tradeoff: stronger sharpening often amplifies artifacts that denoising hasn’t fully corrected. The timing—an unusually large one-shot release of new models—suggests Topaz is betting that users want more control and less guesswork in the pipeline.
Equally important: these models don’t exist in isolation. Topaz announced the release across multiple apps, which signals a broader strategy—enhancement models as reusable engines embedded in different product experiences. If you work with both RAW-to-JPEG workflows and creative editing sessions, model consistency is a workflow advantage: you can standardize “what looks right” across different stages instead of re-tuning settings per app.
Cross-format momentum: SDR-to-HDR and the move toward end-to-end pipelines
Topaz’s broader expansion theme didn’t stop with images. On May 7, 2026, it announced an “Expansion Update” that included a new SDR-to-HDR video upscaler called Hyperion 2, alongside expanded access to newer models across both desktop and web. This is a big deal for AI image enhancers because it reflects a common engineering direction: the enhancement stack is becoming multi-format and multi-surface.
In practice, creators rarely live in a single format. A content creator may capture video, derive stills, and later publish in HDR-like displays or platform-specific color spaces. An SDR-to-HDR upscaler improves tonal mapping and detail perception in moving frames; that same underlying philosophy—repairing the “missing fidelity” rather than just magnifying pixels—feeds back into how image enhancement is expected to behave. The line between image and video enhancement is thinning, especially for users who want consistent results across deliverables.
Topaz also positioned its model access as something that expands across platforms. That’s the business side of enhancement models: once you have trained models, the bottleneck becomes distribution—making them available where the creator actually works. Web access, in particular, changes the value proposition. If a user can run enhancements without a heavy desktop workflow, the barrier to trying—and standardizing—new models drops.
Vmake’s Old Photo Restoration: the consumer test case for “real recovery”
While Topaz is expanding a professional-grade toolbox, Vmake is emphasizing a different metric: can AI restore memories enough that users feel the improvement immediately? On May 6, 2026, Vmake updated its Image Enhancer with Old Photo Restoration features designed to repair faded, damaged, and low-quality photos. The promise isn’t just improvement; it’s repair—implying workflows that go beyond denoising and upscaling into color correction, defect handling, and reconstruction of degraded content.
What makes this noteworthy is the target audience: an “ahead of Mother’s Day” push is a strong clue that the product is optimized for emotional before-and-after transformations. In this market segment, users rarely care whether sharpening is technically “less artifact-prone.” They care whether skin tones look natural, whether backgrounds regain readable detail, and whether the final image feels like the memory they remember—even if it’s not pixel-identical. That requirement pushes developers to balance realism with aesthetics.
Restoration is also where AI enhancers are most likely to be judged harshly. If an image is badly faded, colorization and contrast rebuilding can drift into uncanny territory. If the photo is physically damaged (scratches, tears), the repair stage needs careful handling to avoid smearing. By explicitly framing its feature set as “Old Photo Restoration,” Vmake is acknowledging that modern enhancers are evaluated on recovery plausibility—not just resolution.
What these launches mean for the next AI image enhancer: quality control, not magic
Taken together, these announcements show that the competitive center of gravity for AI image enhancers is shifting from “bigger models” to “more complete workflows.” Topaz is adding specialized image functions (notably sharpening and denoising) in a major April 28 release, then extending the enhancement concept across video and color workflows in early May with Hyperion 2. Vmake, meanwhile, is testing whether general-purpose enhancement can translate into believable restoration experiences through an Old Photo Restoration capability promoted just before a major seasonal gifting moment.
So what should creators do now? First, stop treating enhancement as one button. If your tool offers separate steps—denoise first, then sharpen—you’ll often get cleaner edges and fewer artifacts than if you let a single model handle everything at once. Second, think in categories of damage: noise, blur, color cast, and physical degradation each call for different assumptions. Third, validate output on the hardest area of your image. For professionals, that’s hair, shadows, and textured surfaces; for restorations, it’s faces and clothing where color plausibility is judged instantly.
Finally, keep an eye on platform access. Topaz’s emphasis on expanded access across desktop and web suggests that model availability will become part of the product’s real competitive advantage. In the coming cycle, the best AI image enhancers won’t just be the ones that look good once—they’ll be the ones that help you reproduce “good” across devices, formats, and time.
Key takeaways and forward-looking bets
If you want the practical takeaway from these releases, it’s this: enhancement is turning into reconstruction. Topaz’s Next-Gen model expansion on April 28, 2026 indicates deeper, more specialized control for sharpening and denoising, while its May 7 update with Hyperion 2 and broader web/desktop access highlights the push toward cross-format consistency. Vmake’s May 6 Old Photo Restoration update shows that consumer-grade restoration is becoming a must-have use case, not a niche feature.
Next steps for users are simple: choose tools that separate key enhancement tasks, run tests on the parts of your image most likely to break (edges, skin tones, shadows), and prioritize platforms that let you keep the same enhancement approach across workflows. If these trends continue, the “AI image enhancer” category will increasingly resemble a smart restoration studio—capable of repair, not just beautification.
