* I borrowed a good quality VHS player with SCART connector because it sends RGB in separate channels, improving quality considerably. Don't use the single channel composite video.
* Then I bought a cheap SCART to HDMI convertor and used a borrowed HDMI capture card.
* I recorded it with OBS studio and the resulting video looks very good.
So my total costs were about 20$ (for the adapter).
Important archival work - much appreciated! However some links on the page are not working. Also, it seems like the author has made a Web app to make conversion easy, but I don't see a repo link or otherwise a way to access it.
Since the webapp is pretty opinionated to my setup (ie. linking against AVFoundation, using MPS for inference, always capturing an image after import) I didn't originally think it would be that useful to open source. Happy to do so - are you looking to get something specific out of it?
I had a similar set of tapes, and ended up collecting a chain of connectors – firewire cable, firewire to thunderbolt2 adapter, thunderbolt2 to usb-c.
Instead of cobbling together an impressive array of tools though, I just got a trial of Final Cut Pro and pulled out everything with that. You can get what I think is a three month trial? Anyway, it was plenty for this one time effort of digitizing old Hi8 tapes.
I think I did end up using Handbrake to take the raws down to a reasonable size to give to family members, but the raw footage and project files I stuck on a couple of 1TB Sandisk drives to keep in physically separate backup locations.
If you have a lot of Video8 or Hi8 tapes to digitise, get a Digital8 camcorder. It will most likely play them back quite happily and emit DV over its Firewire port. Digital8 is just DV on a different tape!
I still use DV/DVCAM tapes because I like shooting with old cameras, and I capture the same way I have for about 25 years when I used a VX2000 to shoot DV for a commercial digital streaming company that did all sorts of training videos.
Cheap crappy PCIe Firewire card (back in the day it was PCI, but no-one has that now), and dvgrab to get a raw DV stream off tape, then ffmpeg -i dvgrab-001.dv -c copy whateveritscalled.avi to rewrap it in something the editing software can read. These days I use DaVinci Resolve on Linux, in the olden days I used Premiere 5 on Windows 2000.
Even back then I used to capture on Linux and then bring it into Windows 2000 because only Linux had reliable Firewire support.
What is the value for meaning for keeping home photos or videos? Does anyone provide meta tags, detailed information or narration? How is the photo or video meaningful unless you add your detailed memories to it for others to understand?
Most of my tapes did have pretty detailed narration and date overlays written directly to tape. But even without narration I still had luck doing basic event summarization and facial recognition of family members to build the tags.
VHS is a composite signal on the tape itself. Composite for sake of this thread means black and white detail plus color information. S-VHS has higher bandwidth in the luma (detail) but the same limited color bandwidth. And there are two audio standards. But there is no "RGB out."
At a minimum you want a device with S-Video out (it keeps the two signals on separate wires). You also need a time base corrector. These come in two forms. One is line-based sometimes built into DVD players. This is how Jason Scott at Internet Archive does it and it's wrong.
The other form of corrector requires a separate box and corrects each frame in full. Many boxes claim to be time base correctors but are not. They are "synchronizers" or amps. Don't buy until you understand the differences.
There are two time sources (not really clocks) in a VCR. The first is physical tape wobbling and stretching over a head that's spinning far faster than seems possible. Line TBC is a tiny buffer that reconstructs the sync of the luma on each line.
The other timing source is the overall signal sync. A proper TBC reconstructs this overall sync on a frame by frame basis and presents something sane to the capture card. Without it you'll drop frames silently, audio falls out of sync, and all the other crap that happens when you try to watch video older than an iPhone. Consumer video capture is total crap and you won't see it until you try to encode, edit, or watch it on a different device. And then you'll be very confused working back to the original problem.
But follow this careful path where you actually capture a clean, proper signal and feed it into even the cheapest Blackmagic box and you're good.
ChatGPT will walk you through this and seems to know more about proper ffmpeg settings than the developers themselves or 30,000 conflicting StackOverflow messages on the topic.
Awesome to see this! I actually wrote PySceneDetect, was great to see it getting some use here. Would you be willing to share what parameters you were using? I'm curious why the accuracy was so low.
PySceneDetect only uses basic heuristic methods right now so it does require some degree of tuning to get things working for certain data sets. Your post inspired me to look into maybe integrating TransNetV2 as a detector in the future!
I'm still sitting on some 8mm film that my grandpa shot, and even some of the commercials he worked on. Still trying to find a reasonable way to digitize them.
Also multiple boxes of slides, I know you can buy the scanners (I had one). Wish there was a cheaper automated way hahaha
14 comments
[ 4.1 ms ] story [ 37.9 ms ] threadIf someone knows a faster and good way pls share.
thanks
* I borrowed a good quality VHS player with SCART connector because it sends RGB in separate channels, improving quality considerably. Don't use the single channel composite video.
* Then I bought a cheap SCART to HDMI convertor and used a borrowed HDMI capture card.
* I recorded it with OBS studio and the resulting video looks very good.
So my total costs were about 20$ (for the adapter).
Instead of cobbling together an impressive array of tools though, I just got a trial of Final Cut Pro and pulled out everything with that. You can get what I think is a three month trial? Anyway, it was plenty for this one time effort of digitizing old Hi8 tapes.
I think I did end up using Handbrake to take the raws down to a reasonable size to give to family members, but the raw footage and project files I stuck on a couple of 1TB Sandisk drives to keep in physically separate backup locations.
https://news.ycombinator.com/item?id=24839848
I still use DV/DVCAM tapes because I like shooting with old cameras, and I capture the same way I have for about 25 years when I used a VX2000 to shoot DV for a commercial digital streaming company that did all sorts of training videos.
Cheap crappy PCIe Firewire card (back in the day it was PCI, but no-one has that now), and dvgrab to get a raw DV stream off tape, then ffmpeg -i dvgrab-001.dv -c copy whateveritscalled.avi to rewrap it in something the editing software can read. These days I use DaVinci Resolve on Linux, in the olden days I used Premiere 5 on Windows 2000.
Even back then I used to capture on Linux and then bring it into Windows 2000 because only Linux had reliable Firewire support.
It wouldn't have been crazy to spend $4,000 to have the professionals do it, so long as they produced a reasonably equivalent high-quality result.
VHS is a composite signal on the tape itself. Composite for sake of this thread means black and white detail plus color information. S-VHS has higher bandwidth in the luma (detail) but the same limited color bandwidth. And there are two audio standards. But there is no "RGB out."
At a minimum you want a device with S-Video out (it keeps the two signals on separate wires). You also need a time base corrector. These come in two forms. One is line-based sometimes built into DVD players. This is how Jason Scott at Internet Archive does it and it's wrong.
The other form of corrector requires a separate box and corrects each frame in full. Many boxes claim to be time base correctors but are not. They are "synchronizers" or amps. Don't buy until you understand the differences.
There are two time sources (not really clocks) in a VCR. The first is physical tape wobbling and stretching over a head that's spinning far faster than seems possible. Line TBC is a tiny buffer that reconstructs the sync of the luma on each line.
The other timing source is the overall signal sync. A proper TBC reconstructs this overall sync on a frame by frame basis and presents something sane to the capture card. Without it you'll drop frames silently, audio falls out of sync, and all the other crap that happens when you try to watch video older than an iPhone. Consumer video capture is total crap and you won't see it until you try to encode, edit, or watch it on a different device. And then you'll be very confused working back to the original problem.
But follow this careful path where you actually capture a clean, proper signal and feed it into even the cheapest Blackmagic box and you're good.
ChatGPT will walk you through this and seems to know more about proper ffmpeg settings than the developers themselves or 30,000 conflicting StackOverflow messages on the topic.
PySceneDetect only uses basic heuristic methods right now so it does require some degree of tuning to get things working for certain data sets. Your post inspired me to look into maybe integrating TransNetV2 as a detector in the future!
Also multiple boxes of slides, I know you can buy the scanners (I had one). Wish there was a cheaper automated way hahaha