What's the highest quality AVI to digitize video to for future editing?

[jbandy1]

I have a bunch of old videos on VHS, DVCAM and MiniDV tapes. I'm looking to get them all digitized into a digital file format for archiving on external hard drives. The files may be edited in the future on different editing platforms, it is most important that it works with adobe premiere pro running on a windows machine. so I was thinking to digitize the tapes to AVI but I'm not sure what codec to use. The AVI's can't be uncompressed either because I wouldn't have enough hard drives to back them up. The format from tape will be all Standard definition at 720x480 or 720x486 29.97 fps.

So what I need to know is what is the highest quality AVI (that's not uncompressed) I can digitize the tapes too? What codec should I use? Data rate?

Thank you for your time.

[hech54]

The AVI's can't be uncompressed either because I wouldn't have enough hard drives to back them up.

Best quality and editable?.....buy more hard drive space....plain and simple.

[zoobie]

then I'd choose dv-avi for your SD
it comes in around 13gb per hour

of course, uncompressed is best

[edDV]

The DVCAM and MiniDV should be archived as DV-AVI. Use WinDV to transfer.

VHS for future edit is more tricky. In order of quality and file size ...

HUFFYUV ~30-40 GB/hr
DV ~12.5 GB/hr
MPeg2 ~ 4-6 GB/hr. (8-9 GB/hr CBR for better quality or if noisy)

[jagabo]

As edDV pointed out, the DV sources should be saved as DV AVI. What's on the video tape is already digital, compressed with a DV codec. You want to transfer that digital data from the camcorder to the computer and store it without changes, just wrapped in an AVI container.

For analog to video conversions (video capture):

Regarding HuffYUV: there's also a multithreaded version of HuffYUV that's a little faster. But disk I/O is likely to be your bottleneck when using HuffYUV.

Another lossless codec is Lagarith. It compresses a little better than HuffYUV but it's slower -- both compressing and decompressing. There's also UT Video Codec. MSU lossless, Fraps, etc.

Motion JPEG was used a lot in the past. It's lossy but pretty transparent at it's highest quality settings. Picvideo MJPEG, ffdshow, etc.

High bitrate MPEG 2 with short GOPs and only I and P frames, or very high bitrate with only I frames is a possibility.

Another thing to worry about: camcorder video is going to be interlaced. Be sure to handle it properly.

[jbandy1]

OK, so from what you have all said, my videos from the Mini-DV and DVCAM tapes should be saved as DV AVI. Which means I will get a AVI file using a DV codec that will be a loss less transfer from the tapes to the DV AVI file. Is that correct? It seems like there are a lot of recommendations for the VHS but what will work the best for use later on? Of course it has to be in a format that I can edit?

[hech54]

OK, so from what you have all said, my videos from the Mini-DV and DVCAM tapes should be saved as DV AVI. Which means I will get a AVI file using a DV codec that will be a loss less transfer from the tapes to the DV AVI file.

With what you are proposing to do....these files you are going to create will be VERY close to "uncompressed" as far as file size goes. These are going to be BIG files.....you know that don't you?

IThe AVI's can't be uncompressed either because I wouldn't have enough hard drives

Don't be fooled by the term "AVI". AVI comes in many flavors....and the ones you are going to create by transferring your MiniDV tapes to your computer will be monsters.

[jagabo]

In addition to the sizes edDV listed for full D1 video (702x480):

Uncompressed YUY2 (or other YUV 4:2:2): ~75 GB/hr
Uncompressed RGB (24 bit): ~112 GB/hr

[edDV]

OK, so from what you have all said, my videos from the Mini-DV and DVCAM tapes should be saved as DV AVI. Which means I will get a AVI file using a DV codec that will be a loss less transfer from the tapes to the DV AVI file. Is that correct?

A DV stream transfer over IEEE-1394 (Firewire) to an AVI file is a bit by bit identical copy of the data on tape. It is first generation ready for editing. A 1TB hard drive will contain about 82 hours of DV video.

It seems like there are a lot of recommendations for the VHS but what will work the best for use later on? Of course it has to be in a format that I can edit?

This is a question you must answer. It comes down to editing convenience vs file size on disk.

Huffyuv or Lagarith will have the highest quality but the highest disk consumption. Huffyuv will be less laggy on the edit timeline. DV is a frame based format ideal for editing but more intraframe compressed vs Huffyuv. MPeg is both intraframe and interframe (GOP) compressed. It is the most lossy and most difficult to edit but also compresses to the smallest file size.

For me, the choice depends on the type of content.

Huffyuv or DV candidates would be VHS/SVHS/Video8/Hi8 family or travel videos plus unique TV captures that aren't available on DVD. These will get full restoration filtering and careful editing.

For general TV captures that will only be cut edited, I'm happy with moderate bitrate (4-6 Mb/s) MPeg2.

In the "old days" (pre-2005) MPeg2 bit rates were limited by DVD capacity. With today's cheap hard drives, there is less reason to worry about pre-edit file compression.

[ronnylov]

Instead of high bitrate MPEG-2 I have started to archive my VHS huffyuv captures as h.264 with bluray compatible encoding settings.
Since I capture PAL I use this setting: http://sites.google.com/site/x264bluray/home/576i-pal
But instead of using 2-pass x264 i use one pass with --crf 13 or sometimes even lower.

The file size will be lower than DV but the quality is still very good.
Then I compress the audio to ac3 stereo and use TSMuxer to create bluray/avchd compatible m2ts files.
I store them as AVCHD on DVD-R and as files on hard disk. One DVD-R gives roughly 1 hour "archival quality" AVCHD from VHS captures.
The same file as huffyuv would be something like 30 GB and here it is 6 times smaller. The bitrate is the same as MPEG-2 DVD but the quality is better.
If I use --vbv-maxrate 15000 --vbv-bufsize 15000 I make sure that the bitrate peaks are low nbough to be able to play this AVCHD disc directly in a bluray player also when burned on DVD media.
I like this format because there seems to be good support in editing software (AVCHD and Bluray is etablished formats now), I get hardware acceleration with my graphics card so playback is smooth.

I think bluray/avchd is a format that is better supported than huffyuv in the future. It is easy to edit (with modern hardware and software), play and store.

[hech54]

The file size will be lower than DV but the quality is still very good.

But since the OP also needs these files to be easily edited later on....your post is pretty much useless.

[jagabo]

25 frame GOPs won't be too hard to edit. Eliminating b-frames would make them even easier.

[magillagorilla]

Yep, save h.264 as your final rendering format. Avoid h.264/x264 as an editing format.

DV is easy to edit and is your middle ground between HUFFYUV and MPEG2. Though, not to complicate things, your VHS is more like YUV 4:2:2 and DV is 4:1:1 or sometimes 4:2:0. It is possible that you may see a "color difference" between DV and HUFFYUV. If you ever noticed the pastell like color from some DV footage, this is why.

If your footage is very valuable to you, go by a couple 1TB HDDs and go with HUFFYUV.

[edDV]

Yep, save h.264 as your final rendering format. Avoid h.264/x264 as an editing format.

DV is easy to edit and is your middle ground between HUFFYUV and MPEG2. Though, not to complicate things, your VHS is more like YUV 4:2:2 and DV is 4:1:1 or sometimes 4:2:0. It is possible that you may see a "color difference" between DV and HUFFYUV. If you ever noticed the pastell like color from some DV footage, this is why.

Actually, VHS standard is analog 3MHz luminanace and 0.5 MHz "color under" chroma components. In digital sampling terms, VHS would be adequately sampled at 2.25 : 0.74 : 0.74 (where 4 = 13.5 Ms/s).

4:2:2 has maximum Nyquist bandpass of 6.75 MHz luma and 3.375MHz for chroma components.

4:1:1 DV has Nyquist bandpass of 6.75 MHz luma and 1.69 MHz for chroma components. This is more than enough sampling for VHS/SVHS/Video8/Hi8.

4:2:0 (DV PAL, DVD, ATSC, DVB, etc) has SD bandpass of 6.75 MHz luma and 3.375MHz for horizontal chroma components but is half sampled for vertical chroma.

As for levels, that has nothing to do with sampling but is about setting proper black and white levels. TV or camcorder capture should be at Rec.601 16-235 levels not 0-255. Almost all capture cards respect Rec.601 levels. All HDTV sets and computer video players expect 16-235 luma scaling for either SD or HD.

[jagabo]

Translated: VHS resolution runs about 360x480 for the luma channel 40x480 (yes, that's a 40) for the chroma channels. DV's 4:1:1 is sufficient for VHS capture: 720x480 for the luma, 180x480 for the chroma. Actual measurement of VHS chroma resolution:

https://forum.videohelp.com/threads/319420-Who-uses-a-DVD-recorder-as-a-line-TBC-and-wh...=1#post1981589

Be sure to use a DV decoder that smooths the chroma output when upsampling to 4:2:2 (YUY2 or similar) or 4:4:4 (RGB). Otherwise you get obvious vertical bars of color:

https://forum.videohelp.com/threads/337152-What-happened-to-the-orange-cone-Vertical-Li...ds-and-oranges

[magillagorilla]

you guys are smart

[ronnylov]

The file size will be lower than DV but the quality is still very good.

But since the OP also needs these files to be easily edited later on....your post is pretty much useless.

Yes, and these files are actually easy to edit since most video editing programs supports AVCHD and Bluray nowadays. In worst case I can always convert them back to huffyuv avi before editing if the format would be the issue. Normal final quality for x264 is something like crf 18 to 22 and here I use crf 12 to 13 to get a better quality than normal quality. I could not see any differences at all when reencoding the crf 13 file to a final crf 18 file after editing compared to reencode the original huffyuv file after editing. This was even if watching the video frame by frame. Try it and make your own conclusions.

I want to keep "originals" even after editing if I want to redo the editing in future when better filters are available. But I don't think it is necessery to archive them completely lossless. It is possible to use a lossy compression but choose a relative high bitrate and get quality "good enough" for future editing.

[jagabo]

Another possible problem with h.264 encoding is that it uses 4:2:0 chroma subsampling. If you're only going to do simply cut/paste editing with a few transistions that not a problem -- most delivery formats are also 4:2:0. But if you're going to filter heavily later on that could become an issue -- colors won't be as sharp as a 4:2:2 capture.

[ronnylov]

Another possible problem with h.264 encoding is that it uses 4:2:0 chroma subsampling. If you're only going to do simply cut/paste editing with a few transistions that not a problem -- most delivery formats are also 4:2:0. But if you're going to filter heavily later on that could become an issue -- colors won't be as sharp as a 4:2:2 capture.

Yes but you will get this issue also with MPEG-2 which was recommended in this thread. Since I use this only for VHS captures I don't think colors are that sharp anyway. I also do most processing in avisynth and most filters requires YV12.

Using DV is also an alternative as previously mentioned but then you must convert top-field-first interlacing (which is what my analogue capture card records) to lower-field-first. Not a big issue but it must still be done. Quality is OK but not perfect.

Using huffyuv or other lossless codec: Quality-wise best solution. But it requires something like 30 GB per hour.
You can also use lossless h.264. With x264 use the command option --qp 0. But file size is still very large.

MPEG-2: Lossy but at something like 15 Mbit/s it will be OK (you might need to alter quantizer matrices to avoid saturate the encoder).

Lossy h.264: You get the same quality as 15 Mbit/s MPEG-2 at 8 Mbit/s or lower. Requires more CPU-power for encoding and decoding.

Pick whatever you want. It may be worth using more hard drive space with the lossless codec to ensure best quality and avoid using time for extra convertions. MPEG-2 may be a better choice than h.264 with using a slower computer.

[jagabo]

Using DV is also an alternative as previously mentioned but then you must convert top-field-first interlacing (which is what my analogue capture card records) to lower-field-first. Not a big issue but it must still be done. Quality is OK but not perfect.

Throw away the first field and recombine fields. Or shift the frame up or down by one scanline. Both give perfect results (except the missing first field or missing top/bottom scanline) with YUV 4:2:2 video.

You can also use lossless h.264. With x264 use the command option --qp 0. But file size is still very large.

It won't be truly lossless because a 4:2:2 source will be converted to 4:2:0. Some people recommend using a point resize filter to double the frame height first. But that increases the file size and you have to remember to reduce the frame size later.

[budwzr]

The file size will be lower than DV but the quality is still very good.

But since the OP also needs these files to be easily edited later on....your post is pretty much useless.

The same thing happened to me in another thread recently. I stated a caveat (noun: a warning or proviso of specific stipulations, conditions, or limitations), and a poster chimed in with a solution for something else, then others missed my caveat too, and I ended up as the odd man out.

So yeah, thanks for pointing that out.

[jagabo]

Except h.264 isn't necessarily hard to edit. You can use short GOPs and no b-frames. Or even all I frames and lossless if you want.