Topaz Labs Video Enhance AI Tutorial 

"A few minute of processing" 22 hours later.. still waiting

I agree this product is a very useful video tool. Part of the work I do is to take SD 720x576 video and scanned 8mm/ Super 8mm files, and polish the media for HD documentaries. The best way I've found to do this prior to using Topaz Lab's Video Enhance, is to rigorously prepare the SD files. This helps not only to ensure a better looking SD file for up-scaling, but it will also produce better re-scaled assets for video compositing. Firstly I use ReVision FX's Deinterlace filter to remove the majority of "combing" or stair-casing in SD video. Secondly, I use ReVision FX's directional blur vertically at 1-2 pixels, to remove the remainder of combing. Thirdly, I do a de-noise pass using Neat Video. Finally, I do an initial colour correct of the file and render. Now the prepared SD file is ready to be up-scaled in Video Enhance. This may seem like a convoluted process. But then, what standard do you want for your video work?

It is bloody amazing! I am running the trial version on a slaved PC over ethernet. The Dell tower has an i7 and GTX 1080 GPU and I am controling it from my Mac using Microsoft Desktop Remote. I am seeing a frame every .02 seconds rendered. The results are unreal, it is as good as Topoaz claim. This is the holy grail for video, I have kept all my NTSC videos from 30 odd years ago in case this ever came to pass. I never believed it would but I refused to give up a small hope. It is literally magic.

It would be helpful to have the AI models explained in detail. The help text is just not enough to make a decision to use. Perhaps a blog article or documentation would be helpful to us understanding what each model will do for us.

I've just started using it. I am amazed at where tech has come. Imagine what we can do in 10 years from now.

As a non destructive approach I really like the options it has. Resolving detail out of noise still amazes me, their new updated version must be way better.

I have this software and I can relate it works I tested it whit a 360p video from my 2010 Nokia and the final result was pretty impressive, this is for real IT WORKS

One thing I’ve noticed in some lower quality videos is that they have a SLIGHT choppy appearance in scenes with a lot of motion after the upscale. Does anyone know of any software or Adobe Premiere tricks that can remedy this?

it would be nice a tutorial with the explanation about what Ai mode is better? how can ai know the difference between them? thanx

All good. but what about horizontal line artefacts in shadow after Ehchance work? How to fix it or when you fix it?

Would it be possible to update new features, like adjusting sound output quality? Adjusting crop? Output Field Order / FPS? Some interlaced content are actually meant for 24fps output but sourced as 30 due to 3:2 pull down.

A useful tutorial. At last! Thank you :)

Are there tools to transform lo-fi audio (e.g. from wax cylinders) into hi-fi audio using AI models?

I'm very happy with both,gigapixel and video ai but working with a lot of old 240x320 videos from VHS tapes and manage to upscale many of them to 960's in relatively decent quality but most end up with a dry/flat look and i was wondering if there is any way to add some kind of glossiness to them?Thanx.

hello, does this software allows to make basic video processing like premiere elements to cutout unwanted parts of the videos,... ? is it recommended for mp4 coming from an iphone 8 ? which software in iphone is recommended for a good video before applying your software (native camera of iphone, pro camera, lightroom mobile, camera+...) ? is mp4 ok or do you advise new format heic... ?

best software i have ever had in my 26 years of life on this earth. i converted 480p videos into 1080p its amazing what this software can do.

Do you plan to add a plugin to Premiere Pro? Also, I did not see any great results of original vs after processing warrant to buying this product. But then again, I am really confused on the AI model selection and wonder which one is better. Finally, I tried several passes to see if that worked but again, could not really see great enough improvements to approve such purchase. IS there plans to continue to improve or even add new AI models in the future? I feel this product is a work-in process.....

Works mainly with videos that were well filmed from the start. Almost useless for video8 or VHS yet

Studying the old man comparisons, it looks like a Gaussian blur has been applied on the unprocessed clip to make it look less sharp. I also think enhancing time is prohibitively long

There should be a pause or continue later option for rendering

Topaz VEAI is undoubtedly the best product of its type I have ever used. I've just uploaded to yt a restoration of part of a low budget 1969 film whose source was a DVD which contained a scan from the original Technicolor intermediate. It used the new Iris model on Topaz Video Enhance V3.3.3, and the upscale is from 576p to 1440p. I also changed the speed from the 25fps of the DVD to 23.976 outside Topaz beforehand without transcoding before applying the Apollo interpolation model to change the frame rate to 60fps. I found it was important to avoid any processing or recoding whatsoever of the video file before it was fed into Topaz, as otherwise the quality of the end result would be affected. The video took 22 hours to process on a GTX 1660 Ti, and the upload size for this twenty-minute H265 video was 7GB. Throughout the wrapper used was mkv, with a lossless rewrapping to mp4 before uploading to yt. It's possible to watch the end result on yt using the 1440p setting (note - yt does reduce the quality of the playback slightly compared to what was uploaded due to a lowered bitrate) - ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-N827Qoyb85g.html

Good job, Topaz. You got yourself a new customer. I will buy and have a try!

How do you add audio back into the video after converting the framerate from 24 to 60?

I'm lowkey considering getting this software, I have a 4K upscale of a movie from my childhood but some scenes like close ups and zoomed out shots that get blurred up for some reason, and I would like to apply a "detail preserve" AI filter like the one Adobe After Effects has but without having to pay for a subscription. Love how Topaz is just a one time purchase and you get all the features show on in this video. 😊❤

If your video was shot in 10bit color, will it maintain that in the upscaled video?

I processed my video but the output (mp4) does not play , just black screen. any help?

Which AI Model works best for anime upscale?

Awesome! I have bought it and it works very well! :)

Great tutorial, I will try this software!

It looks suprisingly simple to use, I am not technical minded, I am downloading the trial atm, I wonder how long it would take to do a video around 55 mins. My Pc is windows 11 and 24gb ram.

For a faster processing, witch specs a PC must have? (I got a i5 2nd generation and a video of 8min took me 18 hours)

i'm having a video with the model out of focus. What "enhance" model should i use to bring back some details? Theia Fidelity? Sharpen ? Or maybe Motion Deblur?

My video have 720p and 25 fps, the quality is good but i wish add more details and also reupscale to 1080, any sugget of what ia use?

Can you enhance a complete edit or only individual raw files?

Hi, I love your software so much! I have a suggestion. I would love a feature that enables you to shut your computer down after the softwares done with its purpose. Sometimes a video will take a large amount of time to render or enhance. So this would be a great feature because you can set this setting to shut your computer off when the action is done. Thanks!

how long will it take for a blurry full movie to upscale to 1080 or 2k?

is this better than gigapixel? or any difference between them?

Dear Topaz Labs. I try to uspcale my video (11 minutes) from HD to 4K. They take me 2 days 30 minutes, when Pinnacles 20 take me 15minutes. Please tell me what I doTopaz Video wrong . Thanks advandce

amazing software... but can you please change the default output to a video format? there have been countless times I had to restart an upscale because of a engine error and then realizing half way though a render that I just now have a folder with 100s of jpgs

Does this help reducing flickering for AI videos ? Thank you

Hi. Dose it work with raw video, like 2K raw up to 4K raw?

Do you have to login to the website for the program to work?

Just bought it, totally impressed so far anyway. :)

After I hit "Start Processing" , it starts to load the AI model and then I receive an error message stating "Cannot Write Video - The following video cannot be opened for write. This may be due to no write permissions on the folder, or no file path. Please help :( I made sure the file has permission for writing.

Which model to use to not ruin the out of focus regions of the video? Enhancing models tend to extract details in out of focus regions and it is mostly Ok EXCEPT if there is a person there - it starts to look very bad and unusable.

I really like the results, but I can't imagine workflow process. I have a bunch of 1-2mins videos, and I need just short 2-3sec clips from them. So first I have to cut it in Premiere, save in lossy format, then enhance in Topaz Video AI, save in lossy format, then import again to Premiere and after color grading and other stuff save in lossy format? Would be nice to have a Premiere plugin to avoid unnecessary quality loss through the editing process..

I notice Gigapixel has a model which performs face recovery. However is there something like that in Video? Problem is you have let's say 480p video which got MPEG compressed and where a face which was only 8-16 pixels in size is severely damaged. When upscaling it in vicinity of 1080p, it looks weird and unnatural, makes a weird expression because the upscale becomes too sensitive to source sub pixel error, it creates something uncanny, often looks like a creepy mannequin. So it would need a special tool to restore the natural appearance of faces in the video.

I tried this thing on noisy compressed videos and only Artemis LQ makes a notticeable difference, smoothens a lot but removes noise and blocks. Others don`t do anything at all but simpe upscaling to whatever resolution I choose. Theia and Gaia turned out to be useless. Sometimes with SD videos even Aretimis does nothing. Maybe I don`t understand something.

Is it possible to make the upscaling “smarter” by supplying a specific still shot from the video that you have in higher quality? For example, suppose I’m working on a standard definition version of a clip shot on film, but was also supplied a few frames that were scanned in high resolution off of production stills or whatever. Is that possible?

when i do this the video output is always unplayable. windows says it cant read it or it is corrupt. is there any solution to this?

How to setup if my video is with lower light?

Can it also do photos?

Anyone know of any good alternatives?

I downloaded the trial version but I can't get the audio to work, despite clicking on it in the frame. Technical support can't solve the problem regarding why my movies can't get sound. Also, how do you go to "full screen" so that the frame round the video disappears completely? When I click "view" in the frame and then select "full screen" it's still not full screen. How do I make it so that only the video shows on the screen and none of the frame around it?

Ive seen a guy upscale to 8k with your software from 320p, does this do this as i have not seen your videos show this beyond 4k?

is it possible to crop a video and enhance only cropped part?

I'm currently processing a 20 year old 30 min home video (640x480). I chose "Clean up old interlaced footage", upscaled to 1080p and applied the crop. The processing time is 80 hours. WOW. And that's with a fast Xeon 5160 proc. At 100% scale (no upscaling or cropping) It's 22 hours.

this may sound stupid but where do the upscaled videos go??? i checked in the settings and there's no folder selection.

What happens if you double stack it?

best to use 2560x1080 its semi 1440 it looks almost the same but renders faster the scale I used for semi 1440 is 150-200 for a 720p input vid and 130 for a 1080 when it's a vid bars will be on side has to be cropped in another program to get rid of I first read that gaia is the best but only works good for hd but cant be used all the time cause will sometimes be to sharp or give a plastic type look so the vids I did do I rescaled em but with the art model and they come out more sharp but the thing with the art is it works best set on low but for some vids it gives the eyes a weird looking bugged look or like they have a water effect around em so when some vids have that weird effect on low u have to use medium art but then that causes artifacts so then the only other choice is theia but it's not that sharp as art but it fixes that weird art problem even the some vids that have that weird effect around the eyes the whole vid looks way better then orginal but even theia looks way better still then the orginal even if theia is not sharp as art proteus can be used on any resolution if using a vid where it looks weird using the art model on low or medium use pro instead and adjust settings if still want a sharp vid when using pro I usually use these setting I set to auto and put noise to 20 and compression on 20

Can you please upload a tutorial video where you actually using a few seconds of a real bad quality video like Frozen Ghost-End of The Line and what your software can do to improve it as i'm not into improving slightly blurred ultra-high quality videos that can be sharpened back to its 100% UHQ status with any sharpening freeware.I can not decide to buy the veai because all the tutorial videos use high quality videos and pix which is not hard to bring back to uhq.

Wow. found out about this program while google searching and downloaded the trial. While not perfect it's pretty mind-blowing what it can do. I would only need it for personal use so wait till a sale. Also, love how its not subscription.

I think AI is the only way to get old self-made homevidos on super 8 or even vhs tape to HD quality in the future. I would really appreciate that possibility and look forward to work with it!

IIt's great that it's faster to 1.7.0. However, there are some points of concern. The bit rate after video conversion will be very large, and the file size will also be huge. Is this a bug? This sentence was made by Google Translate.

Topaz actually does work!!! my favorite part is how it enhances and upscales from 1080 to 4k

It helps to use a quantum computer to run the software.

Is there any community WAG or Telegram Group or FB Group for existing customers to have discussions and Q&A? Customers can help each others and suggest features or report bugs.

So can you apply multiple AI effects on 1 video?

I have NVIDIA GeForce GTX 1650 Processor and the rendering for a video takes way too long

Ayuda!. ERROR: Unable to write output. Please make sure that. Your output file has write access. Output video size < 8k (you can set the output as a sequence of images instead). ....No se como arreglar el error.

have a windows 8.1 version?

How does it work with old VHS content? Any examples???

Best to take whatever video you have and process it as far as you can with your non-linear editor before up-resing. Going up too much gives a posterized look to the media. Running 2080ti black, and a liquid cooled 36-core overclocked set up. Doesn’t really utilize all that much of the GPU or CPU. Maybe 50%. Upgrades constantly change what tools are available for manual settings. Really lacks documentation as to each version and why things have been changed. For example some of the tools in this video are no longer available such as a sharpen adjustment.

Please could you add a time code in the frame selection, The amount of times I want to find a point in a film or long video, If I know the scene starts at 0;59;43, How do I interpret that timestamp in the video enhancer haha. It Takes me 10 minutes of searching in frustration, if anyone has an answer?

Yeah is awesome but sometime i lost the audio quality but the upscaled is awesome

Help I can't choose the ai model? It's just blank? Also it gets stuck on the "initializing ai engine"

Well yes, but the space ship visuals (above) were created in a high end 3D animation programme - probably 4K- and Topaz Labs is working with a compressed copy of this original file. Further to my comments below, where Topaz Labs' Video Enhance is a welcome addition to video post production, it should be seen as a product in its early days. I'm using it on a Mac Pro 2019 with lots of grunt, and the render time for Tiff image sequences is still incredibly slow (many hours). Moreover, there are individual plugins some end users would prefer to use by a variety of software vendors, that exclusively denoise, sharpen and up/ down-scale. This is to say the traditional way to prepare video is to use a variety of plugins or apps, as one app rarely will do the whole job effectively. Video Enhance boasts impressive advertising visuals. However, the end result of re-scaling SD video may not be quite as superb as Topaz Labs' advertising suggests. In the end it all depends the source quality (camera lens, resolution etc) of the media. But then, does this surprise anyone?

I'd love to do this to 80's music videos

Does it work on mobile?

many people dont understand something very simple.. this is for bad quality or noisy or compressed or low resolution videos... if you wanna remasterize a video like 1080p with 60fps with very poor compression,,, topaz its not needed unless you wanna go higher to 4k (need very very powerfull pc )but remaster a 4k with 60fps.. topaz got nothing to do.. this is for 360, 480 and maybe 720p low resolution videos and medium to high compressed with noise, unsteady and more.. hope this be usefull

I get many processing errors. Maybe I'm doing something wrong, but I don't know what it could be.

I have a GTX 980Ti and I get like 1-2FPS. Does this not support the AMD GPUs in Macbooks? This program isn't fast, but it is faster than encoding AV1 and whatnot. No problem on my end.

Can not say I have much experience with producing video, but it's starting to come (my customers' wishes) - this looks very promising but I would like a smart function for *image stabilization as well* - then I buy immediately!

Update from my post 2 years ago. Now using a Mac Studio Ultra and rendering to Pro-Res where possible, the speed is staggering.

is the gigapixel only for photos?

you know what I would love? an AI Audio upscale. take for example 8khz 8bit audio, upscaled to 44100 16 bit, and reconstruct detail of sound using a mix of physics to calculate your overtone series reconstruction, and deep learning. Would be cool to reconstruct old recordings to sound more modern, remove pops clicks and drop-outs. the AI could even take text input of what it's supposed to be saying.. and have support for multiple languages. I would buy that

can i upscale to 3440x1440?

For me, a 4min video in 1080p takes up to 5 HOURS to process to 4k (even on my high end video editing PC!). Is that normal? Can I speed up the process some how? Thanks in advance!

Personally I choose Gaia-HD as well, for easier and better upscaling in my opinion. I surely can't wait for the interlaced videos support with hopes for a de-interlacer for them. A little idea/suggestion if y'all don't mind, maybe you guys can add a stabilizing option to stabilize shaky videos and make an object stay at the exact position as an example~

For maximum Quality i tried to export all frames from a mp4 to jpg, send these pictures to photo AI and combine them again together with PS to a new mp4, after i will dub the audio with premiere on it back. But: can you simple add to video enhanced AI a "Topaz Photo AI Frame per Frame" Filter? That could be easier for me :) Thank you.

I thought that this was a fake thing on facebook. It looks really good!

Can I use a asus 15.6 laptop to download topaz video app

Hi there @topazlabs Will you be having an option to Upscale video to 16K 15360 × 8640 On Video Enhance AI??

It’s great definitely i can see the improvement but it’s very slow it takes 6 to 8 hours upscale to video

Who is the actor (old man) in the video?

I know its not just me when I see 58 thumbs up for comments like 22 hours later still waiting ! Personally it took me like 3 hours for a 10 sec clip on an up to date computer. THAT is just NOT gonna cut it for any Serious type projects, I Don't have 4 months to process a 15 min video

I bought this software today I am not seeing any gaia options?

stuck on loading AI model. any way I can fix it?

Why this program is so slow,does this requires fast ethernet or internet or because of 8gb ram??? Please help

Going to try this with an hour-long, originally 360p video. Wish me luck.