Topaz Gigapixel AI 1.3.5 — Mac version

Topaz Gigapixel AI is a desktop app that uses machine learning to upscale images while holding onto the detail that other enlargers turn to mush.

A legacy release that keeps older installs working

Nothing much, and that's the story. This release simply keeps the lights on for the legacy Gigapixel AI line—the one Topaz later replaced with the newer Topaz Gigapixel. You still get the same core job: feeding it a blocky portrait or a noisy scan and watching it rebuild a cleaner, bigger version. The Russian notes point out it can strip away pixelation and compression crud, and it'll batch-process a whole folder from the desktop without you babysitting it.

Separate Intel and ARM installers, English only

The installer still comes as separate builds for ARM Macs and Intel Macs—no universal binary here, so grab the right one. Interface language is English only, and it needs macOS 12 or later. If you've used it before, you'll recognise the workflow straight away. Drag an image in, pick the scale, and let it churn. It's a single-purpose tool that hasn't changed its habits, just its version number.

Dragging an 80-pixel thumbnail and expecting a billboard

Most new users grab the topaz gigapixel ai download mac package, fire it up and immediately drag in a tiny 80-pixel thumbnail expecting a billboard. That's the first mistake. Gigapixel AI isn't magic—it needs some source detail to work with. A postage-stamp selfie won't suddenly sprout eyelashes that weren't there. Start with the largest original you've got, even if it's still small. A 400-pixel headshot will upscale far better than a 100-pixel one.

Another common slip: ignoring the AI model selector. The app ships with different models tuned for different jobs, and leaving it on the default every time means you're missing out. Got a pencil sketch or a logo? Switch to the Lines model. Working with a grainy portrait? The Face Recovery option does the heavy lifting. The model dropdown sits right in the main panel, but newcomers often tunnel-vision on the scale slider and forget it exists.

Batch-processing eagerness trips people up too. Ticking a hundred files and walking away works, but only if you've tested settings on one image first. Nothing worse than returning an hour later to find every result looks over-sharpened because the noise reduction was cranked too high. Do a single test pass, lock in your settings, then queue the folder.

Apple Silicon support lands for the legacy line

Since the jump to version 1.3.5, the day-to-day experience hasn't shifted much—but there are a couple of practical tweaks worth knowing. The biggest is the Apple Silicon support. If you grabbed a Topaz Gigapixel AI torrent or package for an M1 or M2 Mac, the app now runs natively instead of through Rosetta. That means a batch of twenty portraits finishes noticeably quicker than on the old Intel-only builds. The fans stay quieter too.

Another quiet change: the activation process. The package now includes activation bundled in, so you're not wrestling with trial screens or watermarks after installing. For anyone who's used older releases where the activation window kept popping back up, this is a genuine quality-of-life win. The Russian-language notes confirm the same—activation is included, no extra hoops.

Beyond that, version 1.3.5 is a maintenance release. Topaz had already moved development energy to the newer Topaz Gigapixel line, so don't expect fresh AI models or a redesigned interface. What you get is a stable build that runs on macOS 12 and newer, across both architectures, without the compatibility headaches that plagued earlier legacy versions on Apple Silicon.

Test on one portrait, one scan, one screenshot, one line-art

Once the installer's done and you've run the Patch.command file, spend your first session on a handful of test images you know well. Pick one portrait, one old scanned photo, one screenshot with text, and one piece of line art. Run each through its matching AI model and compare the output side-by-side with the original. This gets you a feel for what the sliders actually do before you commit to anything important.

Keep the output format set to TIFF or PNG for the first week. JPEG compression undoes some of the detail you just gained, especially in flat colour areas like skies or skin. You can always export a JPEG copy later. While you're at it, bump the RAM cache setting if your Mac has memory to spare—it speeds up the preview redraws when you're nudging settings around.

If you're working with old family photos, resist the urge to max out the noise reduction. A little grain left in looks natural; scrubbing it completely clean makes faces look waxy. Set noise reduction to around 30, then inch up. And if the app throws a security warning on first launch, head to System Settings > Privacy & Security and click Open Anyway. Standard Mac dance with unsigned apps.

Upscayl is lighter, free, and runs entirely offline

Set Gigapixel AI next to Upscayl and the difference is immediate. Upscayl is free, open-source and runs entirely offline using its own set of AI models. It's lighter on system resources and the interface is simpler—pick a model, pick a scale, go. But Gigapixel AI's Face Recovery model pulls ahead on portraits, and the batch workflow is more polished. Upscayl feels like a capable understudy; Gigapixel AI feels like the specialist who's done the same job for years.

Then there's HitPaw Photo Enhancer, which targets the same crowd of photographers tired of soft blow-ups. HitPaw puts more emphasis on one-click fixes and ships with separate face and colour models too. In daily use, Gigapixel AI gives you finer control over the sharpening and noise-reduction sliders. HitPaw leans toward auto-everything, which suits quick jobs but frustrates when the algorithm guesses wrong. On a tricky scan with mixed content—say an old magazine page with photos and text—Gigapixel's per-image model switching gives you a better shot at a clean result.

Against Adobe Lightroom's Super Resolution, the comparison shifts. Lightroom's tool lives inside a workflow you might already use, so there's no export-import dance. But it's limited to a 2× linear scale and only works on raw files. Gigapixel AI handles JPEGs, PNGs, TIFFs and can push to 6× or beyond. If you're already in Lightroom, Super Resolution handles the quick doubling jobs. For anything gnarlier—low-res scans, screenshots, heavily compressed web images—Gigapixel AI earns its spot on the dock. The topaz gigapixel torrent route puts it on your Mac without the subscription strings attached to Adobe's ecosystem.

• macOS 12 or later
• Separate builds for ARM (Apple Silicon) and Intel Macs