Late 80's early 90's hi8 video upscale

I recently pulled out the old late 80’s and 90’s Hi8 tapes out of the closet and have been getting all the hardware together to get them transferred to digital. I’m using OBS to get them on my PC. I’ve transferred a few short clips to test and used Topaz to upscale them and it’s really done a great job.
I’ve just randomly used some of the included templates and they’ve come out pretty darn good… maybe a little overly processed in some situations.
My questions is I really don’t know where to start… I’ve looked for info but are there any templates specifically for just cleaning up… little sharpening/smoothing/a little anti shaking and upscaling a bit templates for this purpose? THere’s just so many darn settings to mess with.
Also I’m currently using my desktop i9 9900k/rtx4090 rig and it’s taking quite a while for even the shorter clips… likely i’m just overkilling it for what I’m wanting. Or maybe I’m choosing the wrong encoder. I also just bought a Macbook Air M3 16gb variant… are there any hardware encoders that would work with that to make things faster.
Also I know I’m throwing a lot of questions up here but when I buy the license will I be able to use it between my macbook and my windows desktop? Or will it only work on one?
thanks for any info.
I’m attaching a playlist of my unconverted videos along with theyr’e upscaled versions… I’d appreciate any feedback. I just want to get it dialed in so once I get the probably 20 hours of video captured from the hi8 tapes i can just choose my dialed in template in Topaz and start getting them upscaled.

I’ll tell you what works pretty well for my Video8 files. it is not perfect, but plausible results and it retains the original look and feel of the video with not much destruction of details.

  1. I Deinterlace my source video(s) using Hybrid free tool (QTGMC deinterlacing filter) and export as Lossless Compression .
    If your video/source is not interlaced, you can skip this step.
    You can use MediaInfo tool to determine if your video is interlaced or progressive scan (not Interlaced)

  2. Feed the deinterlaced video(s) into TVAI and enhance using Nyx v3 at “Relative to Auto”, leaving everything at default “0” and reducing Dehalo to “-100” as shown: Here
    Enhance another copy of the original video, but this time use Iris LQ v2 in “Manual” mode leaving everything at “0”, as Shown: Here.
    Export both at the same/original resolution (no Upscaling) and Export using FFV1 Lossless compression.
    You can also try exporting a 3rd copy using Proteus v4 with the same settings like you did with Nyx (“Relative to Auto” → Dehalo= -100, the rest at “0”. again, same resolution at Lossless compression FFV1).
    My Videos are very short so I don’t mind exporting it fully few time, but if your videos are longer you might want to export like a 30s Preview each, to save time. Just make sure you export the same preview (part of the video) on all 2 or 3 exports, that way you can truly evaluate the difference in quality once you will upscale them.

  3. Upscale x2 the previously enhanced videos (or previews if you elected to do previews instead) in TVAI (again), this time I use Dione TV , both my Nyx and Iris (2 videos in total, if you also exported a copy with Proteus v4, then 3 videos in total).
    Export using FFV1 lossless

  4. Then I use the Video-Compare Tool to compare the results (Example) between the Nyx → Dione TV vs. Iris LQ (v2) → Dione TV and Proteus v4 → Dione TV, in case you exported Proteus as well.

  5. The winner! then is fed into Handbrake encoding it to x265 to reduce the file size (you can use x264 if you need legacy device compatibility) with the settings of CRF between 21-23.
    That way I get almost as lossless quality but much much smaller file size.
    You can read more about the CRF value settings Here that might work best for you.

EDIT: Lately I discovered that Upscaling x2 Directly from the Original using Proteus v4 works pretty well, with the settings of: “Manual Mode”, Fix Compression=40, Improve Details=40, the rest I leave at “0” (Example), that works well if your video is noisy.
Another settings I found works pretty good for me for Direct x2 Upscale with Proteus v4, is: “Relative to Auto”, Dehalo= “-100”, the rest at “0” (example)
Use the Video-Compare Tool to compare the results…

Good Luck. let me know how it went…

1 Like

@Akila has made some good model recommendations. Additionally, by default, your license can be used on two computers at the same time with the initial user license and we refer to those as seats.
Also, your desktop will have faster speeds than the Macbook Air. You can check the benchmarking page to see results from other users with similar systems.

@Akila Useful tip when doing multi-pass with FFv1 as a lossless intermediary codec is to keep in Topaz’s native RGB48 pixel format, rgb colorspace (full range, by definition) until the final export back to H.264/YUV. It avoids repeated conversion and resampling on multi-pass between YUV and TVAI’s native RGB48le. Once it is in rgb48le and RGB colorspace, you may as well remain in rgb48le for subsequent runs until the final export.

The same goes for any preprocess/deinterlace in FFmpeg or Hybrid into FFv1. You may as well go straight for FFv1 level 3 RGB48le from the start.

I added a preset for video-encoders.json for FFv1 rgb48le.

    "id": "ffv1-rgb48le",
    "encoder": "FFV1",
    "profile": "rgb48le",
    "ffmpegOpts": "-codec:v ffv1 -level:v 3 -slices:v 4 -slicecrc:v 1 -pix_fmt:v rgb48le -coder:v range_tab -g:v 1",
    "ext": [

It is a minor tweak to a multi-pass workflow, but I dislike repeated, unnecessary conversions between YUV and RGB.

I don’t think Topaz uses RGB, not during while exporting at least.

    "id": "ffv1-8bit-420",
    "encoder": "FFV1",
    "profile": "8-bit 4:2:0",
    "ffmpegOpts": "-level 3 -c:v ffv1 -pix_fmt yuv420p -slices 4 -slicecrc 1 -g 1",
    "ext": [

Yeah, it doesn’t include an RGB48le encode preset by default in video-encoders.json, but you can add one to ensure that intermediary lossless codecs maintain that pixel format and colorspace - otherwise you bounce back and forth between pixel formats and colorspaces on each pass.

Converting from RGB48le > 8-bit yuv420p and back to RGB48le on a second pass is, mathematically, a lossy process because it is being resampled down to a lower scale (losing info) and then reampled back up to a higher scale. And therefore it is beneficial that any intermediary lossless codec like FFv1 should preserve the original pixel format and colorspace for subsequent passes to avoid the lossy resampling during intermediary transcodes.

Here’s the proof of the output of tvai_up enhance filter performing a native conversion to fmt:rgb48le, full-range colorspace, in GBR colorspace.

$ ffmpeg -f lavfi -i smptebars,tvai_up,trim=end_frame=1,showinfo -c:v copy -f null -

[Parsed_showinfo_3 @ 0x600003a7cb00]fmt:rgb48le cl:unspecified sar:1/1 s:320x240 i:P iskey:1 type:I

[Parsed_showinfo_3 @ 0x600003a7cb00] color_range:pc color_space:gbr color_primaries:unknown color_trc:unknown

Adding a rgb48le preset for FFv1 allows your intermediary encode to maintain the pixel format for the intermediary files.

1 Like

Hey, excellent info from everybody. Really appreciate it! At the moment I’m about a 3rd of the way getting them ripped to my drive. I found out I have 21 60/120 minute tapes… lots of them! Once I get that done my plan is to take a few short segments of a few different instances… night/day/inside outside and run through some of these recommendations… and see what I come up with. I’ve done a few more random trial and error clips just to see the potential. This software is pretty amazing… definitely very substantial differences in just messing around. I have some late 90/early2000 mini dv stuff that I’m looking forward to messing with too. My first introduction to Topaz was some star trek ds9/voyager restoration and was pretty blown away.
thanks again everyone… I’ll follow up at some point with progress.

hey… so I tried that upscaling x2 is working really well , also some stabilization and adding a bit of grain is looking really good. Another quick question… I’m going from composite off the camera to an hdmi converter… then going from the hdmi to a hdmi capture card. Is this converting it to progressive automatically or do I still select the source as interlaced?

I am not sure, but I believe it will capture it as interlace.
the best way to transfer your videos would be through a Firewire on a digital camera by loading your analogue tapes. i found this much better quality video transfer then capture cards. the better quality you have of a source file, the better success rate you have enhancing them.

1 Like

Interesting. I don’t do it for all footage, but I don’t mind spending huge amount of time getting the most precious footage to the highest quality possible. So you pick the best of the two (or three). Could combining them in post possible yield better results? It shouldn’t. But if you shift between then dynamically, shifting the weighting between tracks … based on a curve, or even manually changing shifting weighting as scenes change. This is a lot of work, but I am talking about getting the best output humanly possible for a few special video clips.

I “digitized” my VHS and VHC-C and hi8 back in the late 90’s by running it through the video in on my Digital8 camcorder. Probably not ideal, but at least it stopped the analog aging process.

I will do some reasearch to learn if I would gain much quality dusting off the old tapes, and digizing them again. If I do, I will try to digitize the same footage multiple times, and then do some type of “difference” layer to see how similar they are. If different, it could be worthwhile combining the multiple takes … as any analog error in playback would be averaged out, possible reducing noise and wobble etc. Would only do this for small fraction of my videos … but if it is worthwhile, the process can be automated (except the digitizing of course).

rca < s-video < firewire, there is no doubt, digitize in firewire, rca and s-video = mpeg with compression, firewire = avi dv without compression

that is what I meant, but, over the firewire port of your Digital8 camera to a firewire port on your PC.
Not over the S-Video -to-> capture card on PC.
You can Use Sony “PlayMemories Home” program to capture it from your Digital8 Camera and it will save it as AVI file(s) DVCAM with low compression…

That is an example of my Video8 tape I captured using my Digital8 Camera over Firewire port.
It is saved as DVCAM.


That is an example of my MiniDV I captured using my MiniDV Camera over Firewire port.
It is saved as DV.


That is an example of my Digital8 tape I captured using my Digital8 Camera over Firewire port.
it is saved as DVCPRO