Topaz Video AI 1.7.1 Mac

Old DVD rips, grainy low-light clips, and choppy handheld footage all share the same problem: the video is watchable, but it looks nothing like what modern displays expect. Topaz Video AI, the Mac app formerly known as Video Enhance AI, tackles that gap with models trained for upscaling, denoising, sharpening, and frame-rate changes.

• Converts SD footage to HD, including DVD, DSLR, and game capture sources.
• Cleans up low-light noise and adds sharpening without a manual filter stack.
• Generates new frames to change frame rate, which helps smooth out handheld or older clips.
• Runs on Apple Silicon and Intel 64-bit Macs, with macOS 11 or newer required.
• Uses a short workflow that the developer describes as a few button presses.

As of version 1.7.1, the interface is English-only. The app is built for people who want a direct path from rough source material to a cleaner export, with film content, commercials, music videos, and YouTube uploads among the use cases Topaz Labs lists.

Archival workers digitizing family footage see the biggest payoff

Not every Mac user needs a dedicated AI upscaler. The people who get the most mileage out of Topaz Video AI tend to fall into a few clear camps.

Archival and restoration workers see the biggest payoff. Someone digitizing family VHS tapes or old MiniDV recordings often faces footage that looks soft and noisy on a 4K display. The app's upscaling and denoising models attack both problems in one pass, which saves hours of stacking traditional filters in a timeline. DVD collectors ripping their shelves to a media server run into the same issue at standard-definition resolutions, and the SD-to-HD conversion puts those files back in watchable shape.

Independent filmmakers and videographers working with DSLR or mirrorless cameras get a different advantage. A shoot that ran too dark or a lens that missed critical focus doesn't always mean a reshoot. The denoising and focus-restoration tools can pull usable frames from footage that would otherwise hit the cutting-room floor. For run-and-gun shooters who can't control every lighting setup, that's a genuine safety net.

Content creators uploading to YouTube and social platforms form a third group. Game capture at 1080p looks dated next to native 4K uploads, and the app's upscaling gives those clips a competitive sheen without recapturing everything. Music video editors and commercial post-production teams use it for quick cleanup passes before grading, especially when the source includes stock footage of inconsistent quality.

The one group that might find less value is anyone whose source footage is already clean, well-lit, and shot at the delivery resolution. The AI models add processing time, and if there's nothing to fix, the render cycles don't buy much.

Drag a clip onto the dark timeline and the preview pane lights up

Open the app on a MacBook Pro with an M-series chip, and the window lands on a dark interface with a large preview pane and a timeline strip below. The first move is always the same: drag a clip onto the window or use the file picker. A short, noisy clip from a dimly lit concert venue loads in a few seconds, and the preview shows the raw source before any processing.

On the right panel, the app lists available AI models. For this clip, denoising and mild sharpening make sense, so two models get toggled on. The preview updates in sections — Topaz Video AI renders a short preview segment rather than the whole file, which keeps the trial-and-error phase quick. Dragging the preview region along the timeline shows how the models handle different parts of the clip: a close-up with heavy noise, a darker wide shot, a moment where the camera jostled.

Output settings sit below the model list. The preset dropdown offers common targets like 4K UHD or 1080p, with a custom field for exact dimensions. Frame-rate options appear next to that, with the ability to generate new frames for a smoother playback speed. For concert footage shot at 30 fps, converting to 60 fps adds motion interpolation that evens out the handheld shake.

Hitting the export button queues the job. A progress bar fills while the Mac's Neural Engine — or the GPU on an Intel machine — churns through frames. The app writes the processed file to the chosen destination, and a side-by-side comparison window pops up at the end. The denoised version looks cleaner without turning skin tones waxy, and the sharpening tightened the edges of stage lights without adding halos.

Switch to ProRes before your first real render for cleaner edits

The defaults work for a quick test, but a few adjustments make the output better from the start.

Output container and codec. The app defaults to MP4 with H.264, which is fine for quick uploads. For editing, switching to ProRes preserves more detail for grading and cuts. The setting lives in the output preferences, not the main panel, so it's easy to miss on the first pass.

Preview quality. The preview renders at a reduced resolution by default to keep the interface responsive. Bumping it to full quality for the preview segment gives a truer read on what the final export will look like, especially when testing sharpening models that can overshoot at higher resolutions.

Model stacking order. When more than one model is active, the order matters. Denoising before upscaling tends to produce cleaner results than the reverse, because upscaling amplifies noise along with detail. The app applies models in a fixed sequence, so testing a short preview before committing to a long render prevents surprises.

GPU selection. On Intel Macs with a discrete GPU, the preferences let you choose which processor handles the workload. The discrete GPU typically finishes faster, but on a laptop running on battery, switching to the integrated graphics keeps thermals and power draw in check for smaller jobs.

Low-quality previews mask fine artifacts the full export reveals

The preview looks different from the export. This happens most often when the preview quality is set to low or medium. The reduced-resolution preview masks fine artifacts that appear at full scale. Switching the preview to full quality for a short test segment narrows the gap.

Long render times on older Intel machines. The AI models are computationally heavy, and an Intel Mac without a discrete GPU can take hours on a multi-minute clip. Splitting the source into shorter segments and queuing them separately helps, as does closing other apps to free RAM. On Apple Silicon, the Neural Engine handles the load faster, but thermal throttling on fanless MacBook Air models can still slow things down after a few minutes of sustained work.

Motion artifacts after frame-rate conversion. Generating new frames to bump 24 fps to 60 fps sometimes produces warping around fast-moving objects. Reducing the output frame rate to a smaller jump — say, 24 fps to 30 fps — creates fewer artifacts. For clips where the original frame rate is acceptable, turning off frame interpolation entirely avoids the problem.

Patch.command won't open on newer macOS releases. The installer instructions note that Patch.command may throw an error on macOS Golden Gate. Right-clicking the file and selecting Open from the context menu bypasses the Gatekeeper block on the first run. If that doesn't work, checking System Settings under Privacy and Security for a pending approval prompt usually resolves it.

A full restoration workflow: old DVD rip to clean HD

Start with a ripped DVD file, a 480p MPEG-2 clip from a concert documentary shot on early digital cameras. The source has three problems: low resolution, blocky compression artifacts in dark scenes, and a soft focus that makes text on stage banners unreadable.

Load the file into Topaz Video AI and set the output resolution to 1080p. For models, enable Artemis for upscaling — it handles detail reconstruction better than the default for this type of source — and add the denoising model set to medium strength. The compression artifacts register as noise to the AI, so denoising cleans up the blocking without a separate deblocking pass. Flip the preview to full quality and render a ten-second segment that includes a dark stage shot with moving lights.

Review the preview. The text on the banners is now legible, and the dark areas lost the blocky sludge that plagued the original. Skin tones on the performers look slightly too smooth, so dial the denoising strength down to low and re-render the same segment. The second pass keeps more natural texture while still cleaning the artifacts.

For the frame rate, leave it at the original 29.97 fps. The DVD already plays back at the correct speed, and frame interpolation on fast-moving concert footage would risk warping around the light rigs. Set the output codec to ProRes 422 for further editing, or H.264 at a high bitrate if the file is going straight to a library.

Queue the full clip and let it process. On an M2 Mac, the render takes roughly a third of the clip's runtime. The finished file opens in QuickTime looking sharper than the DVD ever did on an old CRT, with clean darks and no visible upscaling halos. From there, the file drops into a video editor for trimming and grading, but the heavy restoration work is already done.

• macOS 11 or newer
• Apple Silicon or Intel 64-bit Mac
• Interface language: English