Our Topaz Gigapixel Review: Excellent Engine, Painful Package
Credit first: Gigapixel is still one of the best faithful upscalers money can buy. Its strength is depth of control. You choose between more than nine specialized models (Standard, High Fidelity, Low Res, Text & Shapes, Art & CG, Recover, the generative Redefine, and a dedicated Face Recovery model), and for large-format print work at 4x to 6x enlargement, reviewers still give it a slight fidelity edge over most rivals. It plugs into Photoshop and Lightroom Classic, processes locally for privacy, and batch mode can auto-pick a model per image.
The recurring complaints are just as consistent. Topaz's own community forum carries long threads about over-sharpened hair, warped irises, and unexplained artifact holes, and Trustpilot reviews describe crashes mid-render and slow processing at volume. The hardware floor is real: 6GB of VRAM minimum (8GB for the generative models), 16 to 24GB of RAM, no Linux support, no VMs, no eGPUs, and Intel Macs cannot run the generative models locally at all. Cloud rendering exists as an escape hatch, but it runs on monthly credits that expire and do not roll over. Add the confusion between Gigapixel, Photo AI, and the Topaz Studio bundles, and buying the right thing has become a research project.
Every Gigapixel Model, Mapped to an Upsampler Model
Gigapixel's nine-model picker is powerful but demands that you learn what each model wants. If you are migrating, this is where each one lands:
| Gigapixel model | Built for | Upsampler equivalent |
|---|---|---|
| Standard / High Fidelity | Clean photos that just need size | GAN |
| Low Resolution / Recover | Compressed, damaged, or tiny sources | SeedVR2, which is strongest exactly on degraded input |
| Text & Shapes | Screenshots, scans, graphics | SeedVR2 or GAN; see the dedicated text upscaling approach below |
| Art & CG | Illustration and rendered content | SeedVR2 or GAN, which preserve line work without repainting |
| Redefine / Wonder (generative) | Inventing detail, creative reconstruction | Flux or Stable Diffusion, cloud-rendered with no VRAM requirement |
| Face Recovery | Restoring small or blurry faces | Face preservation option on Flux and Stable Diffusion, or the free no-signup face restoration tool |
The practical difference is that Upsampler collapses the model quiz into one decision: must the content stay identical (Precise) or should the AI add detail (Creative)? Everything else, including picking the right underlying network for degraded versus clean input, happens automatically. For text-heavy sources the text-in-image upscaling guide covers the details, and the free face restoration tool handles the Face Recovery use case without an account.
Print Sizes and Megapixels: the Numbers That Matter
Gigapixel's core audience prints, so here is the arithmetic most marketing pages skip. Print resolution is pixels divided by physical size; 300 DPI is the standard for close viewing, and large formats viewed from a distance drop to 150 DPI or less without any visible loss:
| Print size | Pixels at 300 DPI | Megapixels |
|---|---|---|
| A4 (8.3 x 11.7 in) | 2480 x 3508 | ~9 MP |
| A3 (11.7 x 16.5 in) | 3508 x 4961 | ~17 MP |
| A2 (16.5 x 23.4 in) | 4961 x 7016 | ~35 MP |
| Poster 24 x 36 in | 7200 x 10800 | ~78 MP |
| A0 (33.1 x 46.8 in), at 150 DPI | 4967 x 7022 | ~35 MP |
The takeaway: Upsampler's 100 megapixel ceiling covers everything up to a 24 x 36 inch poster at full 300 DPI, and gallery-wall formats beyond that are printed at 150 DPI anyway because nobody views them at arm's length. Chasing headline-grabbing maximum resolutions past that point mostly produces larger files, not better prints.
Topaz Ended One-Time Purchases. The Math Changed.
For years, the standard advice was simple: if you wanted faithful upscaling and hated subscriptions, you bought Gigapixel once for $99 and were done. That option ended on October 3, 2025, when Topaz Labs discontinued perpetual licenses. Gigapixel now costs $149 per year paid upfront or $29 month to month, the Pro tier is $499 per year, and the entry plan limits commercial use to businesses under $1M in revenue. Existing license owners keep their software but stop receiving model updates.
That pivot removed Gigapixel's main pricing advantage while keeping its main constraint: everything still renders on your own hardware unless you spend expiring cloud credits. Upsampler now occupies the position Topaz vacated. If you upscale occasionally, a one-time credit pack from $10 sits in your account until you need it, this month or next year. If you upscale constantly, subscriptions start at $15/month and include far more than enlargement.
How Topaz Got Here: Twenty Years in Three Acts
Topaz Labs started in 2005 as a Texas maker of Photoshop plugins, long before the AI wave. Its second act began in 2018 with A.I. Gigapixel, arguably the first mainstream AI upscaler, sold the old-fashioned way: $99 once, updates included. That model built enormous goodwill and a decade of "just buy Gigapixel" recommendations across photography forums. The third act is the one you are living through as a buyer: consolidation into Photo AI and the Studio bundles, the subscription pivot of October 2025, and the pending Adobe acquisition. None of that is unusual for a maturing software company, but it means the product's famous virtues, pay once and own it, run locally, stay independent, are each either gone or in question. Evaluate the tool that exists now, not the one the forum posts from 2021 remember.
Adobe Is Buying Topaz Labs
In June 2026 Adobe announced it is acquiring Topaz Labs, with the deal expected to close later in the year. Adobe says the standalone apps will continue and the models will feed Firefly and Creative Cloud, where Topaz upscaling already powers the Generative Upscale feature on Firefly credits. Maybe that works out well. But if you have watched other acquired tools get folded into Creative Cloud pricing, betting your upscaling workflow on a desktop app in the middle of an acquisition is a real risk. An independent web platform does not carry it.
Same Fidelity, No Local GPU
Gigapixel's reputation rests on faithful enlargement, and any honest alternative has to match that first. P Image Upscale, SeedVR2, and GAN are Upsampler's fidelity engines: it raises resolution up to 100 megapixels, removes noise and compression artifacts, and treats the original content as untouchable. The difference is where the work happens. Instead of tying up your workstation's GPU for a batch render, processing runs on cloud hardware, identical speed from a gaming rig, an old MacBook, or a phone. Queue a folder of images, close the browser, download the results when they are done.
Where Desktop Software Still Wins
If you work offline, on flights or in a studio without dependable internet, local software is the only option, and Gigapixel remains the best of that category. Its Lightroom and Photoshop plugins also suit photographers who live entirely inside Adobe's ecosystem. And if you already own a legacy perpetual license and its current models satisfy you, keeping it costs nothing. The trade is capability: no creative enhancement, no inpainting, no generation, no API, and full-price renewal every year for model updates.
One Platform Instead of a Software Stack
The workflow argument matters as much as price. A typical Topaz user still needs other tools around it: something to fix artifacts before enlarging, something to retouch afterward, something to generate assets. Upsampler folds those steps into one pipeline. Run a damaged or soft image through face restoration or creative upscaling, remove an unwanted element with Image Edit, then finish with faithful upscaling for print resolution. Photographers preparing large-format work can follow the portrait upscaling guide for genre-specific settings, and developers can automate the whole chain through the upscaling API.
Weighing Other Upscalers Too?
Gigapixel shoppers usually cross-shop the creative and web-based camps as well. We maintain equally detailed comparisons of Magnific AI (the creative benchmark), Clarity AI (the open-source creative route), and waifu2x (the anime and line-art specialist). Reading two of them side by side is the fastest way to see which camp your workload actually belongs in.
