Table of Contents
Last updated on December 20th, 2021
Topaz Video Enhance AI is a video enlarger that allows users to upsize videos up to 8K ultra-high definition. Upscaling standard definition videos to HD enriches the view experience. In this Topaz Video Enhance AI review, I’m going to test some of my HD videos and upsize them to 4K resolution.
Planning to buy Video Enhance AI? Follow this link and use coupon code aperlust for a 15% discount.
Topaz Labs usually creates software, mainly AI-powered, for still images. Video Enhance AI is their first app for video. I’m going to provide a summary of how their latest software works. But first, a few housekeeping items.
Why You Would Need Topaz Video Enhance AI
Unfortunately, technology moves fast, and not all video cameras can record in 1080p resolution or higher. Furthermore, offering viewers with UHD resolution provides a richer viewing experience.
There are 4 common scenarios a person would need to enlarge a video:
1. Display screens are moving to UHD resolutions. If you are a content creator, or just want to show some videos to your friends or family, then UHD will provide a better viewing experience in terms of quality.
2. There is also the SEO benefit of UHD videos. For instance, YouTube users have the option of searching for videos in 4K. Uploading 4K content will increase your video views.
3. Convert old TV shows like Star Trek or Full House to HD. Many older TV shows and movies are in standard definition with interlacing. Upscaling and enhancing them to progressive video will highly improve the quality.
4. Videos downloaded from mobile apps such as Facebook Messenger or WhatsApp have standard definitions.
This is a common issue when you friend records you doing something that is video worthy like a snowboard trick with their phone. And if that friend shares the video with you via WhatsApp, then the video will most likely be SD (due to downsizing to save space and bandwidth) quality even though it was recorded with a smartphone that has 4K recording.
In addition, keep in mind many smartphones aren’t automatically set to 4K recording, usually it’s 1080p.
Of course, you can always ask your friend to send the video via email. But maybe it was an acquaintance that sent it to you and it’s too much to ask for an email send.
Here are some definitions of common display resolutions:
- 8K ultra high definition (UHD) – 7680 x 4320 pixels
- 4K ultra high definition (UHD) – 3840 x 2160 or 4096 x 2160 pixels
- 1080p high definition (HD) – 1920 x 1080 pixels
- 720p high definition (HD) – 1280 x 720 pixels
- standard definition (SD) – usually anything lower than 720p quality
Topaz Video Enhance AI Review
In my test cases, I used old video footage that I recorded using a GoPro Hero 3 Black at 1080p resolution.
The video quality isn’t the greatest but sets a good benchmark.
Also, no editing such as lighting or color corrections were done to the videos other than cropping and upsizing.
The videos were upsized from 1080p to 4K resolution.
I viewed the quality of the original videos and the 4K videos on my 5K iMac at the native resolution of 5120 x 2880 pixels and also at my working scaled resolution of 2048 x 1152 pixels.
The 4K video has a noticeable improvement in detail.
In order to notice the quality of the videos, I recommend pausing the video at different times. You may also have to view the video at full screen.
Note: In order to view video in 4K or higher, you need to have a 4K resolution display or higher.
Topaz Video Enhance AI Processing Models
The processing models have changed since the initial release of Topaz Video Enhance AI. You may have noticed them in the other photo.
As of December 2021, Video Enhance AI has several upsizing models organized into 4 groups:
- Artemis HQ v12 – Upscale or sharpen high-quality input video, reducing motion flicker.
- Artemis MQ v13 – Upscale or enhance medium-quality video with moderate noise or compression artifacts.
- Artemis LQ v13 – Upscale or sharpen low-quality video with high noise or strong compression artifacts.
- Artemis Dehalo v2 – Upscale and enhance medium-quality progressive video containing haloing, moderate noise, or compression artifacts.
- Artemis Strong Dehalo v2 – Upscale and sharpen low-quality video containing haloing, moderate noise, or compression artifacts.
- Artemis Aliased & Moire v10 – Upscale video with haloing or moire patterns. Aliasing is common in computer-generated (CG) content or from line-skipping cameras.
- Dione Interlaced DV v3 – Deinterlace and upscale interlaced videos. Made for interlaced videos such as DV, DVCPRO, DVCAM, 720i, or 1080i sources.
- Dione Interlaced TV v4 – Deinterlace and upscale interlaced TV source. Made for videos captured from analog TV 480i/576i captured from analog TV/camcorder/camera, or DVDs.
- Dione Interlaced Robust v4 – Deinterlace and upscale interlaced videos. Try this when Dione DV/TV fails. For videos with interlaced frames like 480i, 576i, 720i, 1080i.
- Dione Interlaced Dehalo v2 – Deinterlace and upscale interlaced TV source that contains haloing. Made for videos captured from analog TV 480i/576i captured from analog TV/camcorder/camera, or DVDs.
- Dione Interlaced Robust Dehalo v2 – Deinterlace and upscale interlaced TV source that contains haloing. Try this when Dione DV/TV fails. For videos with interlaced frames like 480i, 576i, 720i, 1080i.
- Proteus Fine Tune v3 – Upscale video with multiple parameters for fine-tuning.
- Chronos SloMo v2 – An FPS conversion model that allows you to change the framerate and apply slow motion effects.
- Chronos Fast v3 – An FPS conversion model that allows you to change the framerate and apply slow motion effects. Input video has a fast frame rate.
I’ll explain a couple of the terms for the upscaling or enhancing models that may sound confusing.
Progressive video sources display even and odd scan lines — at the same time — to make up the entire video frame. On the other hand, interlaced video sources display even and odd scan lines separately — one after the other — to make up the entire video frame.
The p in 1080p stands for progressive. NTSC analog television signals are broadcasted in 480i (interlaced).
Performance
I used my iMac 6-core i5, 32 GB RAM, Radeon Pro 570X 4 GB GPU, to test the software.
When I initially reviewed Topaz Video Enhance AI, it was not optimized for Mac.
It took about 5 hours to enlarge and process a 10-second video at 1080p (60 fps) video to 4K. I just ended up letting the upsizing complete overnight.
The reason Topaz Video Enhance AI was slow on Mac is it required an Nvidia GPU which Macs don’t typically have but is readily available on Windows. The software relied on the integrated GPU and eventually the CPU to render the videos on Mac.
The recommended GPU is Nvidia 1080 GTX 8GB or better. You’ll need to make sure you have the latest Nvidia driver installed.
You can view the system requirements here. But to cut the chase, Topaz Video Enhance will require an Nvidia or AMD GPU with a modern CPU. The new M1 chip from Apple is supported.
It actually took me over 24 hours, with my iMac and unsupported AMD GPU to enlarge a 61-second video from 1080p to 4K with version 1.1.1. But this was the past.
Then Video Enhance AI just got better.
With the latest updates to Video Enhance AI v2.2.0 in 2021, Topaz Labs made significant improvements to the software, especially to rendering times. Even better, it now supports AMD GPUs on Mac.
What used to take me hours to upsize a video with Video Enhance AI on my iMac, I can now render them in a few minutes. That’s progress.
If you have a 60-second video at 30 fps, and Video Enhance AI processes at 30 sec/frame, then your video will take 15 hours to enlarge. If the video was recorded at 60 fps, then it would take 30 hours. These are only estimates and can vary widely. Remember, this is on my iMac and not a Windows machine. For comparison and according to Topaz Labs, on a Windows machine with Nvidia GTX 1080, it usually takes 3-4 seconds to process each frame from HD to 8K. Video Enhance AI is far more superior on Windows than macOS.
When Video Enhance AI used to be slow for Mac. Now it’s faster.
I also noticed a very slight color shift from my 1080p to 4K video. You should be able to notice it in the video. It’s not a big deal since the color isn’t super off but you can always adjust it in post which most videographers do anyways.
Price
Topaz Video Enhance AI used to cost $299.99 and included one year of free updates. Now the price is $199.99
An additional 12-month upgrade costs $199.99.
Follow this link and use discount code aperlust for an additional 15% off.
You can try a 30-day, fully operational Video Enhance AI for free without a credit card
Verdict
Topaz Video Enhance AI works. And now it’s faster on Mac. Really fast. Before I could only recommend it for Windows, but it’s now good to go for macOS with AMD and M1 GPUs.
If you’re working on a video project that requires video enlarging with Windows or macOS, then this product is a great fit. It’s the best video ai upscaling software on the market.
The Topaz Video Enhance AI for Mac was only released on March 5, 2020. I expect them to resolve this Nvidia-GPU-only issue since it will decrease revenue. Topaz Labs actually resolved this statement I made in 2020.
However, enlarging videos is a CPU/GPU intensive task, regardless of the operating system or GPU compatibility. Therefore, a mid to high-performance computer will be required. Remember, the software is working pixel by pixel, frame by frame. Furthermore, upsizing video from SD to HD will be faster than HD to 4K.
One license of Topaz Video Enhance AI can be used on both a Mac and Windows computer.
If you end up downloading the software with the 30-day free trial, please let me know its performance in the comments below. It’ll help other consumers.
If you’re looking for an image enlarger, then you should check out Gigapixel AI. It works well on both Mac and Windows. You can read the Gigapizel review here.
Let me know if you still have any questions from my Topaz Video Enhance AI review in the comments below.
Frequently Asked Questions
Topaz Video Enhance AI is a video enlarging software developed by Topaz Labs. Its was first released in March 2020 and currently version 2 is the latest. The software can improve video quality in standard definition videos and upscale them to 8K. Video Enhance AI is available on Windows and Mac, and requires a GPU.
Yes, the software is great at upscaling videos to HD or UHD. Video Enhance AI is probably the best consumer software to enlarge videos from standard definition like 480i to 4K or 8K. It can also upscale interlaced videos that were originally on TV. You can also try the software with a 30-day free trial.
Topaz Labs used artificial intelligence (machine learning) in Topaz Video Enhance AI to determine detail and noise in videos before upscaling them. This allows the software to upscale only the required data needed and not increase noise or reduce detail.
1. Import video(s). Import video by dragging them from the folder into Video Enhance AI.
2. Crop start and endpoints of footage if needed. Use the curly brackets icons to crop the start and endpoints of the video.
3. Select one of the rendering AI models for upsizing. You can either manually select the AI processing model or allow Topaz Enhance AI to recommend the best model
4. Choose output dimensions – usually will be UHD. Select the preferred final size of the video up to 8K.
5. Choose output format. Select the desired final file type – usually will be mpeg4 (.mp4).
6. Click on ‘start processing’ to begin upsizing. The software will take several minutes or hours to finish enlarging the video.
7. Enlarged video completed. The enlarged video will be located in the same source folder.
Had the same experience on my iMac 27 Core i7 with 32 gig ram. Upsized a video form the 1990s shot with a Hi 8 camera, so it was approx 640×480 to 1080p. Took forever but there is some noticeable increase in sharpness, if not resolution. I believe it is more of a sharpening algorithm than a resolution fix. Will experiment further. BTW my fan ran constantly, which I have never even heard on this machine before.
so far on my macbook pro, it is running wicked slow. I took a 2 hour movie that is a favorite of mine (turner bringing up baby colorized) and streamed it into FCP. I then exported it as an MP4, then took this output into topaz video enhance ai as LQ video to 4K. estimated rendering time was 2.5 days. I am on the second day and it is running a bit behind that estimate. My guess is this will take a lot longer. The system fan is running like a bat out of hell. Thank god I have an external cooling fan to help.
I know they say they are going to improve mac support, but guessing they will make that fix not free for us. Instead since we need to purchase a yearly license to receive upgrades, don’t you think they will only upgrade after each license period? LOL
Trust me, it isn’t any better on a PC running Windows 10 Pro. i7 3.5 ghz processor. For a short 15 second video, I gave up after letting it process for over 6 hours. At that point it wasn’t even 1/4 of the way finished. Not ready for consumer use as far as I’m concerned.
Tried it out with version 1.3.8.
A 21 second video from 1080p to 2160p took around 35 minutes.
Configuration:
CPU: i9-9900 @ 3.6 GHz
Memory: 64 GB.
Video card: NVidia Quadro P4000
Windows 10
Obviously if i convert some (4mins) music video from 480p to 1080p, it takes 3-5hrs – 8GBRAM, intel i5/GTX 1050.
Folks…can’t you even read? This thing was designed to be processed by GPUs…your CPUs don’t matter.
I used a Nvidia GPU (1060, 6GB – not the best but it will do for testing) on a Windows 10 machine. I get 5 sec/frame. I tested with converting a 10sec clip from 1080p to 4K at 60fps. Took about 55mins. The match adds up – 10 * 60 * 5 / 60 = 50 mins. The converted video looks awesome. I also tried Gigapixel. I am impressed with the result. However, I must say that for the amount of processing time, converting a picture is totally acceptable but video conversion just take too long. My 10 secs video is equivalent to 600 pictures. However, that is no fault of the product – AI just take that long. If you want shorter processing time, then you should try a different product (FCP, Premiere). You definitely get faster conversion but the quality also suffers. I wanted to upscale my old videos (mostly 1 hr) to 4K but I don’t see how that can happen with current processing time – will take at least 12.5 days for 1 video not too mention the constant fan spinning (was worried that my machine will blow because of the heat).
Build a Windows 10 gaming PC and you’ll be ready to begin. So I did that with a GTX 1080 Ti. It’s way faster than CPU only, but after reading that the new RTX 3080 is 2.5 times faster, I’m trying to get one of those. The queue to get one is so long that maybe sometime next year I’ll be able to get one. As I’m not a gamer, I just have it hooked to a standard monitor and access the PC remotely via VNC. I had to buy a small window air conditioner and run the compressor on low, with the thermostat on max (so it doesn’t shut off), to deliver a continuous 18 deg C air stream to the GPU. The AC just sits on my table, blowing into the GPU, no venting to the outside. The GPU now runs 55 to 58 deg C. It was going over 90 deg C and shutting down the PC… believe it or not, this is a cheaper solution than watercooling. If I get serious about upscaling, I envision 12 or more PCs inside a small room with a 1.5 ton split AC unit keeping it about 20 deg C inside.
I’ve had it for 3 weeks now and finally got one project done- upscaling a DVD 480i letterboxed movie (one unlikely to come out on Bluray or webstreaming). The movie really wasn’t watchable on DVD with so little resolution and interlaced. After deinterlacing with QTGMC, I had a watchable movie, but not enough resolution.
I wanted control of the x264 encoder, plus I needed to do cleanup on this video with Photoshop (dirt and hair), so I have Topaz outputting to .tiff images.
Topaz VEAI does add detail magically, but no where as much as they claim. I was able to get 706x346p to 1116x576p. I tried 720p too but it looked poor. So I got about 166% more resolution using Topaz. I tried all the models, the Gaia-CG model did best, but there’s no CGI in the movie.
I’m experimenting with encoding to a higher than the final resolution and later downscaling to see if that improves quality any. The theory being that giving it a higher goal will make it work harder.
I have several low and high quality 1280x720p videos I want to see if I can make 1920x1080p. 150% bigger. Most are 50 to 60 FPS (and need to be for the fast action content shown). I’m doing tests with 3 minute segments to find the best models and settings. I’m trying to see if there is any pattern to determine which model (and settings) to use based on the video content.
I’m waiting for the folks on Doom9 and Videohelp to start testing this software. I’d have thought they’d be all over it even with only a 150% increase in detail, but not yet. Maybe I need to start some threads.
I am using Topaz VEAI now for 2 weeks and i got a RTX 3090 on a win10 pc with I-10900kf and 32gb 4266 ram.
For a 1h 1080p 30fps video it still takes 25-30hours rendering it to 4K. Meanwhile the pc takes constand 680 watts and i can heat the house with the hot airstream coming out of the case for 25-30h … and even more hours.
But the render results are not just in upscaling amazing. If you just keep the resolution and render a video it enhances it very viseable. Upscvaling results from old SD to 1080p or 4k (like marriage/family dvds or so) will blow you away.
Before the 3090 beast i got the 3070. And i fear now, topaz has not yet fully implemeted that RTX 30×0 cards yet.
I even used the nvidia studio drivers (not for gamers)
Because the 3090 is in topaz only 20-30% faster than the 3070.
Topas 2.2.0 – MacBook Pro (15-inch, 2017) – Big Sur 11.3.1 – Radeon Pro 560 4 GB 16 GB RAM
SD (DV) – HD (1080P) 0,5 – 0,9 sek /frame
BR Bjorn
I get ~0.55 sec /frame with my amd rx 5700 with artemis high quality v12 1080p>2160p. Obviously your overpriced low-end apple hardware will be MUCH slower.
Someone is bitter and doesn’t understand variables. 🤔
I don’t know why some people are getting such awful performance with this software. I downloaded the trial version and I’m getting really good performance with various different Mac and Windows systems.
I have a newer PC running Windows 10 and a Ryzen 5000 series CPU with a pair of RTX 3000 series GPUs. The video files are stored on a SSD. Currently I’m getting 0.8 to 0.12 seconds per frame converting a low resolution video to 1080p. I’m trying another lower resolution video to a 400% upscale and getting 0.22 to 0.25 seconds per frame. Maybe its user error on their part, maybe they are using a slow mechanical hard drive for their video files, or maybe they think a 8-10+ year old system is much more powerful than it really is.
My only complaint of the software is it can’t take advantage of multiple GPUs and doesn’t seem to do a good job of taking advantage of newer GPUs such as the RTX 3000 series. My GPU usage is rather low when upscaling video and improving its quality/details.
I’m building a video with a mix of 1080p and 4k clips. Should I be working in a 4k timeline and upscale the 1080 to 4k individually or work in a 1080 timeline and convert the final project to 4k?
Hi,
I’m not exactly sure of your question. The timelines in VEAI provide the original resolution with the option of viewing upscaled previews in the timeline at the same time. Are you referring to the timeline in a video editor?
Unfortunately not usable software. The processing time is way too long, really unacceptable. Converting 480p video to 1080p, will take you anywhere from 4 to 9 days, depending on the mode you choose. Thats just 1 video, its only a 2gb size vid I’m converting. And this thing destroys your computers fan.
I was just about to purchase this software until I read that it’s a killer running on an iMac 27 with extremely long render times (compared to a Windows PC). GPU and CPU and fans running at full speed for many hours. Just sound like you’ll end up spending more money to fix or repair your iMac with this much stress on the machine.
If you plan on upscaling anything of any length i.e. a film, then the process is estimated to take 7.5 days @ 29.97 fps for a 2 hour movie. Even a low resolution dvd upscale to 1080P @ 1 hour 20 minutes length (with interlacing so doubles the frame rate) is going to take 1 day 22 hours on my GTX1080 machine which I really don’t plan on leaving it on overnight for 8 days AND not being able to do anything else with my pc other than web browse. Gaming is out of the question while this is ongoing so it’s not practical even if it does improve the picture quality. Compare this to say Xmedia Recode for example, yes the results aren’t as good but the same film de-interlaced and upscaled takes 29 minutes…. So unless the processing time comes down dramatically it’s impractical.
Wow! It must a wonderful experience being as important and superior as you believe yourself to be. Why anyone that is not running the latest GPU or CPU is a full 8-10 years behind you. There’s nothing in between there at all? Wow that’s quite a difference between you and the next person down on the list. I’m running a and getting decent rates but I have a friend running a 3000 series processor and 2000 series GPU with an extremely fast WD black 850 NVME getting results as poor as some mentioned here. I had not even considered it was due to him running an antique piece of crap. it’s always a pleasure to get the facts from a self-absorbed know it all like yourself. Thanks for being there as I would have continued to be stumped by the difference in the performance of out 2 machines if not for your valuable insight and knowledge.
Not sure what you are on about Tim as I was simply pointing out that with the minimum GTX1080 processing times for a film length movie are still incredibly long with an example and comparison. Honestly an Nvme hard drive is going to have very little impact. It’s all about the processing either CPU or GPU depending which AI model you use. Unless you have some uber pc for most people the time taken to upscale probably isn’t a viable option.
I found a way to let Topaz run on my notebook so that the fast-driven fan doesn’t bother me. I reduce the processor output with “Battle Encoder Shirasé” for Topaz to 30%. Instead of shutdown, I use the Power Save Mode. This means that the project continues automatically after leaving the Power Save Mode.