AVCHD

[blaku92]

Hey everyone, I just got the Panasonic TM700 camera (AVCHD/28Mbps 1080p 60fps) and I'm excited to finally be working with this format. I've known about AVCHD for a while, but haven't worked with it until now because I was broke for years and only had a Sony MiniDV handycam . I'm looking to find a frame accurate/free method of transcoding the footage to anything else, like lagarith or huffyuv for example. I'll be doing some research -- especially with the 28Mbps option, because it's not part of the AVCHD spec. I'll be reading several guides the next few days, but any advice or links to where I can learn more about Bluray/.MTS workflows are appreciated too. Is anyone else on the org working with AVCHD? Thanks so much!

[Mister Hatt]

m2ts is a funky container because libavformat breaks on it. Pretty much every open tool uses libavformat. Your best bet would be remuxing to matroska with eac3to and using FFMS2, or if you don't mind it being slower to index and seek but want GPU accelerated decoding then Donald Graft's DGNVDec should be able to handle it assuming you have an nvidia card with a VP2 chip. Both methods are frame accurate. FFMS2 can open m2ts via Haali's splitter however it is not frame accurate, you MUST remux. I would not advise using LAGS or huffy on this myself, unless you are 100% sure that you know what colour coefficients it uses and are encoding to the same colourspace.

[blaku92]

I'm surprised that FFMS2 isn't frame accurate. I'll take what I can get though -- I'm not going to learn how to program any time soon. As far as colorspace is concerned, don't many of the avisynth filters require you to change over to YV12 no matter what? Is there an avisynth script you know of, off the top of your head, that would allow me to find out what colorspace my footage was shot in? I gotta brush up on my scripting knowledge soon.... I feel like I'm getting too rusty. I'll have to look into the DGNVDec once I get my new workstation built. eac3to seems really awesome. I'll definitely give that a try sometime. Thanks for the help!

[Mister Hatt]

info() is a mighty useful avisynth instruction FFMS2 is only as frame accurate as the splitter allows it to be. libavformat sucks at m2ts altogether while Haali's splitter can at least handle it, but not seek accurately. FFMS2 uses Haali for all transport streams in windows, otherwise libavformat. I have a half written m2ts splitter and GPU decoder but it's not for avisynth \o/

eac3to is great for demuxing stuff, and I would recommend that and FFMS2 with the muxed mkv. I'm not a big fan of anything neuron2 makes and DGNVDec is no exception.

[Bakadeshi]

a quick google search indicates that ffmpeg seems to be able to handle it. If thats true, and you want to stay away from commandline stuff, you can try AMVbe and see if it can read it (amvbe uses mencoder which uses ffmpeg libraries). If it can(Just drop the file in the input box and click the Test playback button to test), it will convert it to huffyuv (or many other formats) for you.

[Mister Hatt]

He doesn't want to re-encode it. Which is exactly what mencoder does, seeing as it's demuxer is terrible and fails to do -ovc copy properly. There are plenty of GUI's for eac3to anyway, and IIRC tsmuxer can dump to mkv as well. Huffy is a stupid codec, don't use it unless you have a good reason (you don't).

[mirkosp]

IIRC tsmuxer can dump to mkv as well.

Nah, but it can demux, so you demux with tsmuxergui and mux with mmg, no need to use cli if one doesn't want to, even.

Huffy is a stupid codec, don't use it unless you have a good reason (you don't).

Sometimes Lagarith and Lossless x264 are not options, for amvers...

[Bakadeshi]

He doesn't want to re-encode it. Which is exactly what mencoder does, seeing as it's demuxer is terrible and fails to do -ovc copy properly.

I'm looking to find a frame accurate/free method of transcoding the footage to anything else, like lagarith or huffyuv for example.

afaik, you have to reencode it to get it into lagarith or huffyuv ;p

mencoder demuxes fine, remuxing is whats terrible though, which is why you use mkvtools , mp4box, etc to put it back together. -ovc copy will get you a raw video file just fine in most formats i've tested, but I admittedly don't know much about AVCHD.

[blaku92]

Thanks for all the responses. So far, the only way I've been able to successfully transcode my mts files is through avisynth using Directshowsource after I've put all the streams into an MKV container. When I open that MKV file using ffmpegsource, it doesn't decode the stream properly for playback. The TM700 is a new camera, so I'm guessing it's variation of AVCHD is just outside of the libavformat... for now.

Code: Select all

Telecide(order=1,guide=0) 
Decimate(cycle=5)

Also, the script above seems to do a good job with the IVTC on the 23.976 + pulldown footage. The only problem is, once I load the .avs file into virtualdub, and export it, the entire transcoded clip repeats twice. It basically makes my clip twice as long as the original MKV clip. Anyone have any suggestions on how to fix that? Maybe a "delete second half" command? Maybe directshow just reads it that way and there's nothing I can do about it except mark an in and and out point on the scrub bar within virtualdub before exporting.

[Mister Hatt]

How does ffms2 fail on your muxed mkv?

As far as a good script goes, try TFM(mode=1,pp=5,micmatching=2,slow=2,clip2=TDeint(mode=2,type=3)).TDecimate()

Your script length issue sounds like directshow breaking. Again. If it's really problematic and you have a large amount of HDD space, I would just use JMDec or something to dump the thing to interlaced yuv4mpeg and load that with rawsource or ffms2.