Capturing SD content directly to FFV1?

[aramkolt]

Hello all, I've been reading recently that it is possible to capture directly into FFV1 codec which is a lossless codec that takes up roughly 25% less space than Huffyuv. While you could obviously still capture in a different lossless format like huffyuv and then later transcode to FFV1, it'd save a step if you could store directly in FFV1.

What I've seen described is that it is possible with something called Vrecord for Mac or linux. It basically uses FFMpeg to do the encoding and is a GUI to simplify it. It uses decklink SDI capture cards as the actual hardware. Of course, this does require getting your signal into SDI first.

This article describes doing it on Mac with a BMD Decklink card, though I'm not sure if it works with the most modern Macs/OSs since the article is from 2019:

https://digitensions.home.blog/2019/03/18/moving-video-tape-capture-to-an-open-source-workflow/

I haven't tried it myself, but I plan to try it soon assuming the hardware I've got happens to be compatible.

Wondering if others that have tried other FFV1 direct capture methods either on Mac or Windows and if there were any lessons learned, recommended hardware, obvious shortcomings over Huffyuv capture, or things you wish you knew before doing it.

It's also interesting to me because lossless capture doesn't get discussed as much with MacOS.

[Sharc]

Can't you do this with Vdub2 by just selecting the 'ffmpeg/ffv1 lossless codec' instead of using Huffyuv?
(Menu Video->Compression....-> select 'ffmpeg/ffv1 lossless codec' from the list).

[poisondeathray]

Higher risk of dropped frames with anything that is more resource intensive, that's the main reason why it's not used for direct capture

[aramkolt]

Can't you do this with Vdub2 by just selecting the 'ffmpeg/ffv1 lossless codec' instead of using Huffyuv?
(Menu Video->Compression....-> select 'ffmpeg/ffv1 lossless codec' from the list).

Good question. I haven't really messed with Vdub2 enough since I heard it wasn't as recommended as the regular Vdub, but also wondering if that then does cause it to drop frames as others have mentioned? I'll give it a try either way, just curious if anyone else has tried it and what their experience has been. Thanks!

[dellsam34]

I don't capture to FFV1 directly as my hardware does not allow it but I do output to it after de-interlacing and resizing to UHD lossless, HuffYUV produces large files that exceed YouTube's file limit so FFV1 allows me to upload the lossless files to YT by staying under the file size limit.

[rgr]

Can't you do this with Vdub2 by just selecting the 'ffmpeg/ffv1 lossless codec' instead of using Huffyuv?
(Menu Video->Compression....-> select 'ffmpeg/ffv1 lossless codec' from the list).

Good question. I haven't really messed with Vdub2 enough since I heard it wasn't as recommended as the regular Vdub, but also wondering if that then does cause it to drop frames as others have mentioned? I'll give it a try either way, just curious if anyone else has tried it and what their experience has been. Thanks!

FFV1 requires more CPU power. Why not use UT-Video? It's much better for this purpose, even including YUV422P8 BT601 codec.

[poisondeathray]

Even with enormous amount of slack resources ( very low CPU % used, SSD's ) people report more dropped frames with more computationally intensive codecs for whatever reason. It doesn't make any sense: with modern hardware, SD capture, it should be more than ok even with HD capture... but for some reason higher rates of dropped frames occur with more computationally intensive codecs

The FFV1 in vdub uses intra encoding and you don't have control over those or some of the other FFV1 settings. But FFV1 compression ratio excels when you use long GOP (temporal compression), not intra encoding for many types of sources. But keeping 15-60 frames in the buffer for a live capture is going to significantly increase the risk of dropped frames - probably not even possible - that's why should compress it offline separately if the goal was high lossless compression ratio.

UT video is a good choice for direct capture, but compression ratio is worse than say lagarith, or ffv1 intra - and a lot worse in many scenarios compared to using lossless temporal compression

[aramkolt]

Thanks for the info everyone! I was kind of surprised that FFV1 has as much of a risk of dropped frames if it's a suggested capture process for video archives, but could just be that they are using specific hardware that is less likely to do it maybe.

[poisondeathray]

I bet some people do not check too closely, or they assume that it works ok . It's easy to run some tests, align the streams, do some PSNR tests

For archival purposes, archivists still recommend to keep using intra , and not to use any temporal compression. The rationale is if you get 1 corrupt frame, potentially that could affect several frames making them corrupt as well when using long GOP encoding.

The trade off for intra encoding is larger filesizes. For clean material , low motion - FFV1 intra vs. FFV1 long GOP encoding might be 1-1.3x the filesize. But for noisy, high motion material, the benefit of using temporal compression is much less

[dellsam34]

Thanks for the info everyone! I was kind of surprised that FFV1 has as much of a risk of dropped frames if it's a suggested capture process for video archives, but could just be that they are using specific hardware that is less likely to do it maybe.

Have you considered capturing in a different lossless codec to avoid dropping frames then converting to FFV1 for archival?

[aramkolt]

Capturing in a different lossless format and then converting to FFV1 is fine, it would just save a step if there was no downside to direct FFV1 capture which is what I was most curious about with this thread to see if anyone currently does it in practice or not. I think a lot of times, people capture all of their footage first and then get to the editing/viewing part much later, so if the archival format part could be done in one step with the bonus of space savings, that'd be worth looking into, but I understand if the risk of dropped frames or minimal space savings makes it not worth doing.

[dellsam34]

For me it's never a one step, I capture the entire tape including between scenes and start and end static noise, then I have to put the file in Vdub2 to clean it up into one file or several files depends on the contents, and then another step where I de-interlace, crop and/or resize in which I output FFV1.