huffy/lagarith

[GH]

how would i go about converting avi files to huffy or lagarith?

i've had the amvapp downloaded for quite some time. normally i would edit with avi (divx/xvid) footage, which yes; i realize is a bad idea so don't bother post the gif. and since i last took a break from editing i bought an external hd so now i have a place to store the lossless files.

i've never converted to this or have attempted, so i'm unsure of what programs will be needed although i'm almost positive it's all included in the amvapp. the only details i've gathered is; it takes a while, and the files are rather large. this is where i regret not taking the time in learning the technical aspects of editing years back.

which keys in to a few more simple questions:
how long does it take, and how large is the average (25/30min.) file?
do these files take away some of the video preview lag, or add to it?

[GH]

i wish i could edit that post, there's some bad wording in that OP.

[Kariudo]

.avi should be pretty simple. Provided you don't need to do any work on the footage (IVTC, dot crawl/rainbow removal, etc...) you only need virtualdub or virtualdubmod which is, as you guessed, included in the amvapp. Vdubmod is old/outdated but is fine for SD footage.

Huffy and Lagarith themselves aren't really slow, it's when you have to clean up the footage via avisynth that you start to see big slowdowns. I can't really give any encode times atm, but file size should be somewhere around 15GB per episode IIRC. For that reason, it is recommended that you select your clips using vdub/mod and encode those smaller clips losslessly.

Huffyuv is faster to encode than lagarith but ends up taking more space.

Huffy and Lags aren't meant for playback, but for the purpose of previewing in your editing program I'd think you shouldn't run into problems. The lag is probably a result of your computer not being up to the task.

[Mister Hatt]

Filesize depends on pixel count and complexity. From my own experience, a 24 minute episode has roughly the following filesizes with lagarith:

704x400: 4GB
720x480: 7GB (NTSC DVD footage, higher complexity than broadcast or encoded image)
640x480: 5GB
848x480: 7GB
720x540: 7GB
960x540: 9GB
960x720: 10GB
1280x720: 12GB
1440x1080: 19GB
1920x1080: 26GB

Can't say I use huffy enough to know how that goes. If you want raw YV12 footage, the filesize in bytes will be 1.5 * height * width * number of frames. You can figure out the size of an uncompressed RGB image from that.

As to your other questions: Speed of decoding/seeking depends on the amount of data IO and the speed of the decoder/splitter, along with how far spaced I frames are. At anything under 1280x720, both should be instant. Huffy should be instant at anything under 1920x1080, as at that point data IO limits tend to kick in unless you have something like a RAID60 SAN over XFP 10KBase-SR fibre like in the AutoDesk Stone+Wire rigs.

I'm in a good mood: How To Encode A Regular VFW Decodable AVI File To Lagarith Or Huffyuv:

I'm going to assume you have lagarith installed, and ffdshow for huffy because pengvado's sweet ffhuff is far better than regular huffyuv:
1) Open VirtualDub
2) File -> Open, and open your DivX file or w/e
3) Audio -> None
4) Video -> Fast Recompress
5) Video -> Compression
6) Lagarith -> Configure
7) Set colourspace to what you need, leave all three boxes unchecked
8) Press OK a bunch of times
9) File -> Save as AVI

To use ffhuff instead, select ffdshow instead of lagarith, hit configure, select the encoder tab, and set it to ffvhuff, and then change whatever other options you want. I would recommend CABAC instead of CAVLC but everything else is generally ok on defaults.

[Qyot27]

To use ffhuff instead, select ffdshow instead of lagarith, hit configure, select the encoder tab, and set it to ffvhuff, and then change whatever other options you want. I would recommend CABAC instead of CAVLC but everything else is generally ok on defaults.

That's FFV1. ffvhuff doesn't have a coder switch (at least not in ffdshow, anyway), just left/plane/median prediction and the adaptive huffman tables option. @Keish, make sure the adaptive tables option is checked. The prediction mode is mainly just a speed vs. size control - left is fast but huge, plane is a balance, median is smallest of the three but takes longest (although it's still pretty fast).

[Mister Hatt]

Actually it does, but it has been forever since I've used ffdshow and it probably just lacks it being exposed to the interface.

[Qyot27]

Which parameter is it then? Because -coder is ignored when using ffvhuff.

Code: Select all

ffmpeg -i "[Сode] Betwixt and Between.mp4" -vcodec ffvhuff -coder ac -context 1 -pred 2 -vframes 100 test.avi

and

Code: Select all

ffmpeg -i "[Сode] Betwixt and Between.mp4" -vcodec ffvhuff -context 1 -pred 2 -vframes 100 test.avi

came out with exactly the same size (~14.8 MB). Removing the adaptive huffman tables option (-context 1) results in a file ~26.1 MB, though.

[MOD258: Lame fix, but works. ]

[Qyot27]

Haha, I should've known that filename would screw up the bbcode tags.

[mirkosp]

Haha, I should've known that filename would screw up the bbcode tags.

Fixed for you.
Anyway, I think he said just ffhuff too besides ffvhuff... not sure if that's why, but I guess trying won't do harm.

[Mister Hatt]

It seems I was wrong and it got removed. Apparently it was never part of the interface in the first place, only the code underneath for other things to use. My bad~