Aaron, ditto from me. I am also very interested in Mike's progress since I am thinking about getting one myself. I am torn between HV20 (or its successor) and HG10. I have no experience with HDV and AVCHD whatsoever, my camcorder is a regular MiniDV, as simple as a hand saw.

So, to reiterate, I would like to clarify:

* CoreAVC is AVCHD codec, right? Is it decoder only or it can encode back into AVCHD? Does it work seamlessly with media players and NLEs? Or do I need some sort of special support for installable codec in my NLE?

* I use Sony Vegas Movie Studio 6, AFAIK it does not support plugins. Will it work with CoreAVC codec? Also, how does Sony enable/disable certain file types? They say that HDV or AVCHD are not supported by VMS6, but if I had a proper codec, why would VMS care? I don't want to upgrade my NLE, I just want to use a codec to read HDV/AVCHD files.

* What exactly does HDLink do? I understand it is a separate utility.

* Neo HDV is a intermediate format. Again, can it be used with Sony VMS, in particular with VMS6 which does not even support HDV out of the box, or do I need an NLE with explicit support for plugins like full Vegas?

* Mike, what is Quicktime for in try 5?

CoreAVC Pro, which I purchased for $15, is a decoder only. It does not encode. I'll get back to you on whether it works seamlessly with Windows Media Player when I re-GHOST my system for the next "start from scratch" test. As far as NLEs go (Non-Linear Encoder?), I don't know as the only one I have right now is Nero, which seems to always use its own built-in filters.

I can't fully answer this one, as I don't have Vegas. The tips from Les and David about graphedit led me to use the tool to attempt to get Nero Vision working as the NLE for output. The way it works is, you open graphedit, drag the file you want to check into its window (in the case of the HG10 it's a file with a .MTS extension), and graphedit will "draw you a map" of what DirectShow filters will be used by default to translate the video and audio. If you look in the Try 5 .PDF I put together (further up in the thread) you can see that at that point the system was using the CoreAVC codec by default. What makes the CoreAVC codec necessary is that it can be easily configured to NOT deinterlace the fields in the file, which is necessary in order to recover the 24p data contained in the interlaced fields. That may be the other issue with Vegas. Even if it can use the DirectShow default filter (CoreAVC codec) it may not be able to handle the pulldown process itself. That's the process of re-combining the progressive frame pieces that are embedded inside the interlaced fields back together. THIS is what CineForm HDLink does, once it can be fed with a codec that doesn't mess with the original interlacing. The output from CineForm HDLink is a 24 fps progressive file which can now be fed to just about any NLE, and interlacing/deinterlacing is no longer a concern.

I *think* I might have just covered this above, but basically, it is used with the HG10 output, the CoreAVC codec (or another that leaves interlacing alone) and the InterVideo audio codec (which comes with the camcorders software package) to "pull out" the 24p information from the interlaced stream so you wind up with an output file that actually has the progressive frames in it, without "hiding" them in a 60i stream (telecine). HDLink is, of course, not limited to the HG10, but that's my focus here.

Again, I'm probably not the right guy to answer this, but my GUESS would be that Vegas 6 would handle the CineForm HDLink output as long as you have the CineForm codecs on the machine with Vegas. HDLink outputs "easy" 24p files encoded with the CineForm codecs that get wrapped into an .AVI or .MOV file. The output from HDLink is HUGE (due to trying to retain maximum quality through to the final output). Les Dits original 6 MB 4 second test file turned into a 32 MB .AVI after David Newman pulled the 24p out of it and saved it as an .AVI (My own test yielded a 36 MB .AVI). That's ~500 MB PER MINUTE. That's why the Neo HDV output is intermediate, no one can afford to archive that kind of size.

Neo HDV (HDLink) can output to either .AVI or .MOV. Personally, I'm fine with .AVI but didn't know if others would be interested in Quicktime (.MOV) files. One of my next moves was going to be to ask a Pro on here if it mattered (.AVI vs. .MOV). I was basically "kicking the tires" to make sure it worked so I knew whether or not to visit the .AVI vs. .MOV question later. Again, look at the .PDF from Try 5 - on the 1st page you'll see the Neo HDV Prefs sheet where you can see that it's possible to specify either .AVI or .MOV.

All that said, if you're open to .WMV files (don't know if Vegas can work with those), I'm hot on some testing right now using Window Media Encoder instead of CineForm to pull down the 24p from the stream. So far, what I see is promising, I'm hoping to post later about it. I don't think Pros like .WMV files at all, but I'm no Pro (that's for sure).

I started out hoping to find a way to get the camcorder output to a "REAL FULL HD" AVCHD DVD-R or even BD-R/BD-RE or HD DVD-R at TRUE 1920x1080p24. I figured if I could do this, it would justify the purchase of a Playstation 3 which would allow playback of the 1080p24 through HDMI at an actual 1080p24 (instead of dealing with pulldown from an interlaced stream). This is, to me, the "nirvana" of camcorder output (as sad as I know that is!) :)

I'm finding the reality is that I don't even know if this can be done for less than thousands of dollars, so I've now settled into getting the 24p out of the 60i stream and utilizing my Home Theater PC (HTPC) to display the TRUE 1080p24 (be it 1440x1080p24 or 1920x1080p24). This means that, for me, the "intermediate file" would effectively be "the end of the line." I would simply play it back on my HTPC whenever I wanted to view it in 24p. When the authoring tools mature and the High Def format war is over (if it ever is!) I can just take my "intermediate files" and use them to create Blu-ray or HD DVD discs at will. Or use my archive of original AVCHD from the camcorder, assuming the tools will be smart enough at that point to get the 24p out. That's why this is the "cutting edge", eh? At least all my footage of my kids from this point forward would be in "Full HD".

The pluses so far for Windows Media Encoder are that I'm happy with the output at lower bitrates than CineForm (meaning archiving these "intermediate files" is a possibility) and I have the option of scaling the output .WMV to 1920x1080 if I like (which means the display resolution will match the encoded resolution and the display driver will not need to provide scaling to provide the final 1920x1080 output). The big downside is that the encode itself is A LOT slower.

OK, I thought this would be short...back to the encoding tests....

:P

mike@slavis.com

Here are the results...Part 1
 

I'll try to keep this as short as I can. I've been on a crash course in video editing this weekend, and have learned several things (I'll stick to those that are relevant to the discussion here). First, let me finishing answering Michael J's question from above:

Yes, CoreAVC works seamlessly with any media player that supports DirectShow for its filtering (Windows Media Player is the one I'm using). It also works with Windows Media Encoder, which is the one I'm using for my current testing.

OK, I'll try to sum up the good and bad points of what I've found.

The good:
Windows Media Encoder (WME) does seem to support Inverse Telecine (IVTC) on the HG10 .MTS files (with CoreAVC filter utilized - and with no deinterlacing in the CoreAVC filter) to yank out the 24p from the 60i stream. I have experimented with several combinations of CBRs and VBRs and settled on what is optimum for my setup (Pentium D 2.8 GHz). In my case, this is VBR for both video and audio with a quality of 90. The output is very nice.

More good news with WME, you can "leave" the video at 1440x1080 and specify it for anamorphic display, or you can have WME resize it to 1920x1080 and save a square-pixel output file (so your player doesn't have to re-resolve it for Full Screen display).

More good, WME includes a file editor which can be used after encoding to clip sections out of a large .WMV file without the need for re-encoding.

More good, WME is a free download from http://www.microsoft.com/windows/win...Downloads.aspx. To get this working, the only $$ I had to spend outside of the camcorder itself was the $15 for the CoreAVC Pro filter pack. Here's the configuration I used:

1) New install of Windows XP Professional
2) CoreAVC Pro codec pack 1.5.0.1 installed ($15)
3) Corel Apps from Canon HG10 disc installed (Included with camcorder)
4) Microsoft Windows Media Encoder installed (Free download)

More good, the file sizes are small, although if you want the same kind of quality that CineForm outputs, you can specify a greater quality and/or bitrate to achieve it. Just watch it - when I went over Quality 90 my Pentium D 2.8 GHz couldn't keep up trying to play the files back.

The bad:
WME is SSSLLLOOOWWW to encode. It takes about 45 seconds - 1 minute to encode 4 seconds of footage. That means an hour of footage will take approximately 12 - 15 hours to encode (on my Pentium D 2.8 GHz). This may be acceptable to those who do not shoot for hours every day (I don't - I am a "family" user who will, on average, take "a while" to fill up 1 hour of footage) and/or have a separate computer they can use for encoding (I do) and/or don't mind leaving the computer on overnight to chug away (I don't).

More bad, it's a fairly manual process (at least for me right now). I still need to learn more about WME, it seems to have a scripting language, perhaps that can be put to good use. I have been processing individual files for my testing, I'll need to figure out later how to batch-encode multiple files (I'm sure there is a way).

The worst bad (put on your thinking caps for this one, I'll be asking for some assistance at the end of it), the "Auto" setting of the IVTC section in WME does not work 100%. It does a fair job, but after examining thousands of frames for hours this weekend, I can say that it is not 100% (I look for interlace artifacts (combing?) in the output to see how good a job it's doing). I have compared with the original as well as CineForm output.

The "good" side to this is that WME offers 10 manual specifiers for how it should IVTC, and I've been able to successfully isolate "the one" that works successfully for each test case I've tried. The bad side of *this* is that the HG10 is not consistent in how it telecines, and the method used is potentially different for every file, meaning to find "the one" to use I actually need to trial-and-error all 10 (it's been a really fun weekend). :P

I'll try and explain, the pros will know what I'm talking about but the rest might not - I'm still learning myself so hopefully I don't royally mess this up...look at http://en.wikipedia.org/wiki/Telecine#3:2_pulldown for a description of telecine and IVTC. The issue here is that the progressive frames are "trapped" inside the interlaced fields, which play back at a different frame rate. To pull the frames back out from the interlaced fieldscorrectly, you need to know what the starting point is. It could be AA, BB, BC, CD or DD. WME calls these AA Top, BB Top, BC Top, CD Top and DD Top. WME also has settings for AA Bottom, BB Bottom, BC Bottom, CD Bottom and DD Bottom. Pick the right one, and the progressive frames magically "pop out" of the interlaced stream. Then WME re-encodes them at 24 fps and you're back where we should have been at the beginning - 1080p24 (for REAL). Pick the wrong one, and some frames may still come out progressive, but NOT ALL WILL. OK, now here's the question I need help with (I may also ask this as a separate post in a different forum which may be read by more people since it is a general question):

***** Here's the question:
Is there a tool, preferably free, that will examine telecined input and tell you the method used inside it? (AA Top, BB Top, etc...) This knowledge would lead me to always know what IVTC method to choose in WME.

If there isn't, I have another grassroots method I noticed from a pattern in the output files. Using VBR and a constant quality (I used 90 for all my initial test settings) the "correct" IVTC method always had the smallest encoded file size (which makes sense since the progressive frames should be smoother than ones with interlace artifacts in them, and therefore easier to compress). This held true for the 3 tests that I did with VBR Quality 90 (and also the test on Les Dit's file). Will it always be true? Not necessarily, but it's better than nothing. If there's no surefire way to ID the IVTC mode ahead of time, I plan on trying to script WME to encode the 1st 10 seconds (maybe 1 second is enough) of a file with all 10 IVTC keys and then check the file sizes, choose the lowest, and then script WME to process the entire file with the correct key. Thanks, Canon, for making this so easy. :P

Sound like too much work? You're probably right. My testing used graphedit. Open graphedit, drag the file (.MTS, .AVI or .WMV) into it and start stepping through the frames 1 at a time. Look for interlace artifacts (combing). NOTE: DO NOT run 2 graphedits, 1 with .WMV and the other with a CineForm .AVI. When I did this and went frame by frame on the CineForm .AVI it showed interlace artifacts. When I loaded just 1 graphedit for the .AVI the frames were progressive. I definitely can't explain that one, but it had me freaking out for a while. "1 graphedit at a time, please." If you look at the 0011 .WMV 29.97 you'll see combing all over the place (whenenver I moved the cam and even in the numbers changing - look for "stripes" in the black numbers of the LCD). In the 0011 WMV BBTop there is no combing (all frames are progressive). Check the 0011 WMV Auto and you'll see the IVTC doesn't "get all" the frames right. There's also a 0011 WMV 1920, which has square pixels and a 0011 WMV 1920 Quality 100, which is about the max quality you're going to get from WME (My machine can't play it real-time, it's not fast enough). The CineForm .AVI version is out there as well. And the test wouldn't be complete without Les Dits original file. Using my "smallest filesize" detection method, I mark this file as LesDit-24p-1440-Qual90-BCTop-Progressive.wmv (It is 5,665 kZB. The next smallest from the bunch is 11,126 kB). I'm including Quality 90 output from this 1440 anamorphinc run as well as Quality 90 1920 square-pixel output. Les, let me know what you think.

Thanks for listening!

mike@slavis.com

(Keep this short, yeah right. I overstepped the posting limit so will post the files next.)

Part 2
 

Here are the files for you to check out, if you like. Again, these are on my personal FTP server, which is serviced through a cable modem (limited upload speed) so you must be patient. Also, I'll try to keep them alive as long as I can but can't guarantee they won't disappear some time.

YOU SHOULD RIGHT-CLICK AND SAVE THE FILES TO YOUR HDD BEFORE TRYING TO VIEW, Please. :)

My "raw" test file, straight from the camcorder (plus a file name change) in AVCHD format. You can drag this to GraphEdit to make sure you're using the right filter to decode it (in my case, CoreAVC Video Decoder (DEINTERLACE TURNED OFF) and InterVideo Audio Decoder):
ftp://66.66.199.125/Cineform/00011-24p-Tv-F2.4-1000.MTS (10,026 kB)

The test file converted to .WMV without DeInterlacing or IVTC with VBR Quality 90 and left as 1440 anamorphic. You can use this file to see the combing present in the result. NOTE: You DO NOT need CoreAVC (or any additional fitlers, including CineForm) loaded to view this or any .WMVs (you just need a computer with enough CPU power, mine is a Pentium D 2.8 GHz):
ftp://66.66.199.125/Cineform/00011-2...29.97-NoDe.wmv (3,184 kB)

The test file converted to .WMV with BBTop IVTC with VBR Quality 90 and left as 1440 anamorphic. Compare this to the one above and notice that this one has 24 fps (vs. 30) and the interlace combing is gone:
ftp://66.66.199.125/Cineform/00011-2...ROGRESSIVE.wmv (2,260 kB)

The test file converted to .WMV with AUTO IVTC with VBR Quality 90 and left as 1440 anamorphic. Compare this to the one above and notice that this one has interlace combing occasionally (the worst is the stopwatch on frame 5. The rest of the file is pretty good, but again the IVTC wasn't "Perfect" in Auto). Also note the filesize is larger than the "Perfect" BBTop version above:
ftp://66.66.199.125/Cineform/00011-2...-1000-Auto.wmv (2,437 kB)

The test file converted to .WMV with BBTop IVTC with VBR Quality 90 and scaled to 1920x1080 (square pixel output). Compare this to the one 2 above:
ftp://66.66.199.125/Cineform/00011-2...-1000-1920.wmv (2,774 kB)

The test file converted to .WMV with BBTop IVTC and scaled to 1920x1080 (square pixel output) and Quality 100 video/Quality 98 Audio. I cannot play this file smoothly on my machine due to lack of CPU Power, it is basically an "intermediate" file which can be used to feed an NLE (if it supports .WMV) to maintain the highest quality. Compare this to the one above:
ftp://66.66.199.125/Cineform/00011-2...20-Qual100.wmv (23,559 kB)
Here is a .PDF with the details of the WME properties used to create this Quality 100/98, 1920x1080 square-pixel .WMV:
ftp://66.66.199.125/Cineform/00011-2...20-Qual100.pdf (645 kB)
Here is the actual WME settings file, in case it is of use to anyone:
ftp://66.66.199.125/Cineform/00011-2...20-Qual100.wme (11 kB)

The test file converted using CineForm Neo HDV to a CineForm intermediate .AVI. You'll need the CineForm decoders on your machine to view this, I'm including it just as a reference:
ftp://66.66.199.125/Cineform/00011-2...4-1000-001.avi (31,628 kB)

Les Dit's original test file converted to .WMV with BCTop IVTC with VBR Quality 90 and left as 1440 anamorphic. Compare this to the CineForm ones referenced earlier in this thread:
ftp://66.66.199.125/Cineform/LesDit-...rogressive.wmv (5,665 kB)
Here is a .PDF with the details of the WME properties used to create this Quality 90, 1440x1080 anamorphic .WMV:
ftp://66.66.199.125/Cineform/LesDit-1440-Qual90.pdf (657 kB)
Here is the actual WME settings file, in case it is of use to anyone:
ftp://66.66.199.125/Cineform/LesDit-1440-Qual90.wme (11 kB)

Les Dit's original test file converted to .WMV with BCTop IVTC with VBR Quality 90 and scaled to 1920x1080 (square pixel output):
ftp://66.66.199.125/Cineform/LesDit-...rogressive.wmv (7,496 kB)

Thanks again,

mike@slavis.com

Wow. Nice work in chasing that down. Get some sleep, man....

Mike,

Thanks for your detailed investigation. I looked into why you had issues with the CineForm conversion, which is designed to save you all this headache. It turned out your sample image was quite difficult for the automatic pulldown detection to extract on the fly. The CineForm algorithm is designed for live capture from HDMI or HDSDI source, where camera compression doesn't get in the way--if this was a live source you wouldn't have had this issue. Even to a compressed file, under normal shooting conditions done of this would come up (as it doesn't for Les's clip), but now we have a test case we can directly address the issue.

The CineForm Pulldown detection is constantly looking for 24p cadence, even of sequences that have changing cadence the CineForm algorthim will adapt to fine the new cadence. The cadence repeat happens over 5 frames, so the cadence is typically found after 5 sometimes 10 frames, from the start of the clip or the point of change. In your sequence the cadence was found a frame 50, that is long time, although every frame after that is correct. When using the Elecard decoder on the same sequence, the cadence was found on frame 38, still not fast, yet it demonstrates how the source decoder can impact the testing. Once we determined the issue's cause, the cadence was detected on frame 10 (as it should be.)

Here is the sequence : http://www.miscdata.com/downloads/00...00-001-new.avi

Add this is the updated pulldown detection component : http://www.miscdata.com/downloads/CFVideoChange.zip
Unzip add replace this component in C:\Program Files\Common Files\CineForm. Then restart HDLink and convert again. This patch will be standard in the next build.

A bit more info...
 

David,

Thanks for the info, and the updated file. Glad I could help you improve your pulldown detection routine. :) I will test this myself on my test files again when I come back around to that.

Meanwhile, here are a few other items I've learned while continuing to test...

1) DO NOT USE COREAVC 1.6. They "upgraded" from 1.5 to 1.6 a few days ago, I downloaded it to try it out and found there was no longer a "No Deinterlace" option in the codec. Go figure. I actually have 2 tickets in to their tech support, FYI, and since they're relevant to using WME on the HG10 files I figured I'd put the results here.
***********
I purchased the CoreAVC codec to allow me to work with the .MTS files from my Canon HG10 AVCHD camcorder. The codec has been excellent with one problem, which is fairly major (for me). I am currently using Windows Media Encoder 9 to recode the .MTS files to .WMV files (WME9 is using the CoreAVC codec to read the .MTS, verified via graphedit). However, I can not specify a Mark In/Out time in the "Sources" tab after specifying an .MTS AVCHD file as input. It gives me an error "Invalid Pointer" = 0x80004003 whenever I change the Mark In or Out time.

Is this a configuration issue, or something that can be fixed?
----------------
Thank you for writing to the CoreCodec Support Group.
While we would like to help you this feature that you hare having a problem with has nothing to do with the CoreAVC decoder. Sorry for the inconvenience.
*************
Hi, just upgraded from 1.5 to 1.6 and noticed there is no longer an option for No Deinterlacing in the config box. Instead there are Weave, Blend, Bob, and Hardware. Does one of these correspond to NO Deinterlacing? I need no deinterlacing so I can properly IVTC 24p from 60i (Canon HG10 camcorder output).
-----------------
Thank you for writing to the CoreCodec Support Group.

That feature is being re-added in the next version. Sorry for the inconvenience.
**************
2) The files straight from the camcorder always have a "Top" orientation for IVTC. Mostly it seems to be BCTop, but it does vary. However, I've now tested over 15 files and noe of them have ever been "Bottom" oriented. That cuts the possible keys for IVTC down from 10 to 5 (at least for .MTS file straight from the cam), which is a big help.

3) The Ulead software can be used to clip sections from an .MTS file from the camcorder. Tell it you want to create an AVCHD DVD, then import and clip your footage, choose Export Selected Clips -> Fast Export AVCHD Compliant Video. This will save the footage as an .M2TS file, which is a straight clipped "copy" from the original .MTS (no recompression) and retains the 24p in 60i telecine which can be fed to WME for "manual" pulldown detection and encoding. This helps when you have a large "waiting" period at the beginning of the clip and you want to start off with some action. A note on this, I have not extensively tested to see if 2) from above applies. These clipped files may have the pulldown cadence "Bottom" oriented. More testing will confirm/refute this.

4) It has been the case EVERY time that in a section with any motion, the properly IVTC'd .WMV file created from WME has had the smallest file size. This would make it easy for me to batch-drive WME, except for some programming faux pas' on the part of the WME / CoreAVC team. Which brings me to:

5) I can't get WME to batch process my files for proper pulldown detection/IVTC. Here are the factors working against it

5a) CoreAVC Mark In/Out - WME will not accept Mark In/Out when using CoreAVC. In the user interface of WME I am "getting around" this by choosing "Limit by Duration" in the Output tab to test-encode the 1st four seconds of video using the various pulldown keys and eyeballing for the smallest file size, which I then use to encode the entire file. CoreAVC says this is not their problem, although I don't buy it.

5b) Scripting Time (Mark In/Out) - My attempts at using a VBScript and driving the WME script (WMCmd.vbs) to do the various keys on the first 4 seconds of a file, grab the lowest file size output and reencode the entire file using it are failing because the "Time" parameter of the WME Script is failing (it is eguivalent to the Mark In/Out times of the interface) - See 5a). Also, I cannot use my "workaround" from 5a) because the WME programmers did not put an equivalent command for "Limit by Duration" into the scripting engine.

5c) No option to close WME when starting from command line - I tried abandoning scripting and running the WME interface from a command line, but while I can do this like "WMEnc.exe /start MyProfile.WME" (this will use the "Limit by Duration" information saved in the .WME), the encoder stays open after encoding, making it difficult to know when to shut it down and start the next one up. I'm currently stuck here, and looking for ideas.

Thanks,

mike@slavis.com

thanks
 

Mike, wanted to say a Huge thank you for all your work here. Thanks are due to Les Dit also.

I was sitting on the fence about HG10 vs HV20, It's my first camcorder ever and I really wanted to skip tape and go HDD, but not if the 24p stuff was "unusable" as per some sites and also I needed to get it done on a home-user budget (read - can't sneak the cost of Neo HDV past the wife :) so all your work helped me finally take the plunge and I got the HG10 last week. I love it so far and am working on getting WME to encode the files correctly for play on PC.

BTW, I got v1.6 of coreAVCPro (which has no deinterlace option) and am waiting over a week for their tech support to reply to me providing me with a link for v1.5 download so i just tried setting the deinterlace option to "hardware" in v1.6 and WME (when set to deinterlace IVTC with BC Top) seems to have done the job right when testing with Les Dit's original file (I don't see any combing). I have an ATI 9550 graphics card and it's deinterlacing setting is set to "weave" (can't seem to turn it off there either) so am not sure if that is being used but the WMV file created seems to be ok anyway.

Keeping the MTS files as archives and creating WMV files for distribution to friends and family is what i'll be doing for now. Also am going to give authoring HD DVD to regular DVD-Rs a try (http://avsforum.com/avs-vb/showthread.php?t=705146) and then maybe get an XBOX 360 or a cheap HD-DVD player... :)

Mike, any chance I can get the PDF & WME files from your post?