Hello, I just want to know what is the best A.I Model for a amateur VHS video in 720*416 resolution, I want to enhance this in 4k. Does the model Artemis LQ make good result ? I join on this topic screenshot from the original video, thank you very much. !
I would try the gaia cg first, but I can only speak for myself. I know others had good results with artemis too.
Thank for for your anwser, I will try as soon as i could
I’ve only been using this for less then a week. But most of my footage has been VHS media.
So far I haven’t had any real luck with the Artemis renders. However Gaia HQ and CG have given me best results. I prefer CG in most cases but have had a few where HQ did better. Remember you can preview a few rendered frames to view the quality.
I’ve been also trying to enhance VHS footage and have been experimenting with different settings. The results look gorgeous (I’ve used various Dione options primarily but have been testing others, as well), but I’ve consistently been experiencing an issue that I’ve seen come up as a common problem with Enhance AI in google searches - the resulting export featuring occasional skipped and repeated frames, which creates a ‘jittery’ playback effect.
Is this something others have experienced when working with VHS footage? A part of me wondered if it has to do with how Enhance AI deinterlaces.
I’ve asked Topaz support about this and they haven’t been able to pinpoint where the repeated frames issue is coming from so far.
Hi, I use Dione TV on anything interlaced and set it to 400 resolution. I render out as a series of frames at progressive de-interlaced TIF 16 bit and then import these into my Vegas Pro editor for colour correction and adding back the sound. It’s a long process and uses lots of hard drive space for a long video but the results are just awesome. I have just restored two DV tapes that were 18 years old, and have been able to take still frame ‘photos’ of the event which are so good that people who see them are convinced they were taken with a camera. Not done a VHS attempt yet, but you must use a deinterlacing preset to get the best from any interlaced footage. I have never experienced the ‘frame jitter’ problem in over 30 video projects restoring. I wonder if it is to do with the interlaced order. If it is not set correct it will cause jittering, but that would be continuous, albeit only obvious on the moving part of the image. I will try some VHS and see what happens.
Thank you for the feedback!
I’m glad you haven’t experienced that issue.
I’ve mostly been using ‘bottom first’ as the field order. Do you know which you’ve done?
Also, what file type are you going into Enhance AI as, out of curiosity? I’ve been going in there with AVIs. I’m currently trying to run them through Handbrake to deinterlace and/or detelecine and come into Enhance AI with that having been done and also with them as mp4s, but I’d rather not downgrade the quality in Handbrake before going into Enhance.
I just purchased the video enhance ai software and am about to improve my 10 old VHSvideos.
Would you be willing to describe step-by-step (and settings in video enhance ai) how you were able to get great results in restoring your media?
Ive got a rtx 3080ti and currently running my first scale up with deinterlacing, Dione TV option and 400 resolution. Side-by-side comparison look great for now…!
What do you mean with " I render out as a series of frames at progressive de-interlaced TIF 16 bit" ? Sorry i am a newb…
Hi Bjorn, VHS videos are very difficult to restore to anything good, but with Topaz AI there are a lot of ways to get the best.
To start - let me say, if you have a ‘first generation’ VHS tape - that means the tape that was in the camera that shot the video, then that is the best start point. Next is a VHS that has been a copy of some other format, like 8mm or better still Hi-8 or S-VHS or DV tape. These give even better results. That is because their original video is higher detail than VHS, so the loss in making a copy is not so bad.
So - there are two things that VHS does to make the quality fit the old tape when it records. It does half the luminance detail (the black and white part of the video) and even less of the detail in the chrominance (colour part of the video). If you understand this, then it helps to work out better results.
The worst tape, which I have at the moment, is a VHS of a wedding made in 1998 which was a copy of the camera VHS tapes edited. So it has very poor quality and is a good example of how much work can be done.
The first thing is to capture the video to the computer in the best format. This is interlaced AVI usually.
Remember, in addition to the loss of detail, in older videos there is also a lot of noise in the dark part of the video, or if there is not good light on the original video. This noise can make the detail fuzzy and not easily seen.
My first question is, - do you have video edit software on your computer?
I use Vegas Pro as I used to be a Vegas beta tester when it was owned by Sony, and know how good it is for helping to restore video with Topaz Video AI. Also, let me say, that too much sharpening will ruin the picture quality of VHS restoration and make it look visually artificial and uncomfortable. You will understand that my background working for ten years at the BBC has taught me a lot, which I am trying to tell you about in a couple of emails! So, if you follow my ‘exact’’ process, then you will get the same results as I do. BUT - it does depend on having the Vegas video software to complete the best quality final video. I’ll do more information for you when I know that you have Vegas. Also, you need to do this with plenty of hard drive space. So for an hour of video, to get the best, you will need a hard drive space of nearly 3 TB spare - just for the restoration. The Vegas Pro video lets you make HEVC mp4 files of your results that look amazing and also are fast to make because they use the power of the 3080Ti to encode. I think Topaz ‘may’ include that part soon, but it does make a huge improvement in workflow with Vegas. You can also make many different formats and sizes of video with Vegas Pro in all sorts of codecs. More soon…?
tony11, maybe you could make a step by step tutorial for everyone here, public?
I also would like to restore/upscale VHS, but don’t have the capacity of free 3TB… nevertheless I would be very interessted to read about your process with Vegas Pro and Video Enhance AI.
Hiya, yes I’d be delighted to - I’m right in the midst of three simultaneous restoration runs at present, and so it may be a couple of days, as I can use some of the results to show what I did.
One is taken from a live stream, one is from a first generation VHS copy, and another is a restoration of an Iraq mission from the DVD, with the 101st Airborne that an old house share and BBC work friend (sadly died a couple of years ago) participated in as second cameraman on that job So, a wide range of sources that will be converted. VHS is the worst - BUT - you can help by pre-conditioning the media in Vegas Pro. BTW, you do not have to have 3TB if you are happy to render to mp4. This is just my pro approach that gives me a lot more flexibility, but I am doing these projects more and more direct to mp4 …( but still having dropount with audio sometimes) . So, this is where Vegas Pro comes in, because you can sync back the audio easily in Vegas Pro, if you render Topaz AI mp4 without audio.
I deinterlace my 480i video to 480p with a 3rd party program – I get better results when I feed that into VEAI. Then I typically use Artemis Medium Quality and add some grain. Once done, I bring it into FCP and fix things.
If I don’t deinterlace prior – then the results are unpredictable.
The most important thing is a good VCR. Then you need a good time based corrected (TBC). Most importantly you then need a good capture card or device. Capture the video a 720x480 or 720x576 if PAL.
I use a Blackmagic (Teranex) device – which is high end – but it will stabilize your video and even provide you with double the frame rate for smooth video (59.94 or 50 fps).
VHS tapes can look really good or they can look really bad. It depends on the equipment used to record the video and how you capture the video.
720x416? Why use such a non-standard resolution? How is the software going to know what to do with it?
I think people would get more predictable results by using 480i or 480p, 720p, 1080i or 1080p, etc.
First off going from VHS to 4k is being overly optimistic. I’d only ever try going to 1080p or more often than not 720p. From the 2 test pictures the footage doesn’t appear to be in the best of shape to begin with either. I’d try using Dione TV either keeping the same resolution or going to 720p. If you have editing software I’d stick to the original resolution and do any editing/colour correction before any upscaling as VEAI will “enhance” all those scratches/dust. I’d recommend only upscaling once you have de-interlaced and cleaned up the original footage. You will probably still find once upscaled that it has enhanced some dead pixels and other stuff that wasn’t noticeable before. If you are hell bent on going higher than 720p I’d still go to 720p first then upscale using either GAIA which I find gives more natural look without oversharpening or Proteus with some manual fine tuning.
BTW [steven.toushin] if you are applying de-interlacing/detelecine in handbrake then don’t try and do it again in VEAI that is what is most likely to be causing the jitter. Trying to de-interlace footage that has already been made progressive will cause problems to occur.
You may also want to read this about your jitter problem:
Another source of interlaced video: three-two pull down
Sometimes referred to as a 2:3 pulldown, three-two pull down is a process used to convert material from film to an interlaced NTSC display rate. This involves taking content created at 24 frames per second and converting it to 29.97 frames per second, which is the signal frame rate of NTSC video. This process involves duplicating fields, two from one frame and then three from the next frame or the process can also be vice-a-versa. Consequently, it’s common for this to be called 3:2 pulldown or 2:3 pulldown as well, with the numbers used interchangeably to describe the effect.
Reverse telecine: removing the 3:2 pull down
Also known as inverse telecine (IVTC), reverse telecine is a process that can be used to remove the effects of taking a source and stretching it from 24 frames per second to 29.97 frames per second. This involves removing the added information from the frames to return it to the 24 frames per second.
For example, frame 1 might be converted into frame 1A and frame 1B through interlacing, with each being a vertical odd or even sequence that is interlaced. However, frame 2 might be converted into frame 2A, frame 2B and frame 2C, with the last one being duplicated content that is used to gradually increase the frame rate. As part of reverse telecine, this added content would be removed to restore the video to its original frame rate.
If you want to live broadcast content that previously had a 3:2 pull down applied to it, it’s recommended to encode it with a reverse telecine process ahead of time before the broadcast. Apple Compressor and Handbrake, the latter calls this process “detelecine”, are two examples of programs that can be used to achieve this.
If you are doing frame rate conversion with telecine/de-telecine before de-interlacing it will also throw a lot of bugs in.
I agree - get your video into a standard definition format, using quality capture equipment. Your playback equipment will make a huge difference. These cheap video capture USB units are junk.
In my opinion, using software like Handbrake or other non-professional utilities that are free will not give you good results.
You either go professional all the way or you run into an issue. Capturing VHS video is best done with equipment made to actually do it properly.
I bought this software to up-res SD video that was already at its best. But going from 480i to 4K is crazy.
Go in increments - fix what you don’t like and then move up another resolution in my opinion.
The problem here is this is to be a professional tool - expecting a professionally captured video input. Otherwise we always say “Sh$t in → Sh$t Out”
This isn’t magic.
Sorry old post. I was wondering if you were able to make that step by step guide.