AVCHD Footage to SD (AVI for Premiere)

[foochuck]

I shot some footage with a Sony camcorder in 1080i AVCHD format. I'd like to downgrade the footage to Standard Definition - 720 x 480 (Widescreen at 1.2 PAS).

Does anyone recommend any software to do this? I'm going to be editing the footage in Premiere Pro & After Effects CS3.

Thanks!

-foochuck

[Baldrick]

I would use avisynth, like

DirectShowSource("PATH\myfile.m2ts",fps=23.976)

and save as video.avs, then open in for example virtualdub, resize and save as avi lagarith.

Or use a frontend for avisynth like Xvid4PSP. You also need to install codecs so you can decode the mt2s, ffdshow and haali media splitter is usually enough.

But doesn't premiere pro support avchcd? Maybe just latest cs4 version.

[foochuck]

I think just cs4 supports avchd - I don't think CS3 does.

Baldrick, are there any tutorials out there for the workflow you suggested?

[Soopafresh]

https://forum.videohelp.com/topic346331.html

Or Xvid4PSP

or the easiest (and best, but not free) is NeoScene

[creamyhorror]

If you're starting out at interlaced 1080 and you want max quality, you should bob-deinterlace before you resize. (Resizing without deinterlacing damages field structure.)

In this thread, poisondeathray suggested this Avisynth script:

AssumeTFF()
ConvertToYV12(interlaced=true)
interp=nnedi2(field=3) #double rate, starts with top
yadifmod(order=1, field=-1, mode=1, edeint=interp) #TFF,field set to order,double rate
Spline36Resize(720,480)

This is very good deinterlacer, but also very slow. You'll have to install the Yadifmod and NNEDI2 plugins from http://bengal.missouri.edu/~kes25c/.

The output from the script will be progressive 720x480. You can import that into your editor, edit it, and interlace once again if needed. Follow Baldrick's method of installing VirtualDub and the Lagarith codec (you should find a guide).

edit: If you use soopafresh's first method, note that you'll get 848x480 square pixels rather than 720x480 with PAR 1.2. Or you can choose the 1920x1080 output option, which means you retain the full resolution for importing (higher quality but much more taxing for editing).

[poisondeathray]

In this thread, poisondeathray suggested this Avisynth script: .....

If he's starting from AVCHD, it's already YV12 so ConvertToYV12() isn't necessary (as opposed to DV in the other thread, which was YUY2)

foochuck would probably want to re-interlace it as well, assuming the end goal was NTSC DVD

AssumeTFF() 
interp=NNEDI2(field=3) #double rate, starts with top 
YadifMod(order=1, field=-1, mode=1, edeint=interp) #TFF,field set to order,double rate 
Spline36Resize(720,480) 
SeparateFields()
SelectEvery(4,0,3)
Weave()

Another option is to use the Premiere AVS Import plugin, which works with CS3, and you can get practically any file into it that way, including AVCHD.

But the benefit of resizing to an SD lossless intermediate is that editing will be a snap even on old PC's

[creamyhorror]

Ah, I forgot to remove the ConverttoYV12 line.

PDR, have you used NeoScene? What advantages does it have over Avisynth+Ut/other lossless codecs (besides lower bitrate)?

[poisondeathray]

PDR, have you used NeoScene? What advantages does it have over Avisynth+Ut/other lossless codecs (besides lower bitrate)?

No I haven't.

I use UT because it's truly lossless, and decodes fast (I haven't seen a head-to-head test vs. cineform for multithreaded decode speed, but I suspect cineform is faster)

[creamyhorror]

offtopic: I've been testing x264 intra-only as a potential lossy intermediate codec, but it may not decode fast enough in NLEs even then (though it's plenty fast to play and seek with any multithreaded decoder). Even if I use DSS2 via AVS to load the output in Premiere CS3, it doesn't play back smoothly (while Ut does). Not sure what I'm doing wrong...should I be trying DGAVCDec?

[poisondeathray]

offtopic: I've been testing x264 intra-only as a potential lossy intermediate codec, but it may not decode fast enough in NLEs even then (though it's plenty fast to play and seek with any multithreaded decoder). Even if I use DSS2 via AVS to load the output in Premiere CS3, it doesn't play back smoothly (while Ut does). Not sure what I'm doing wrong...should I be trying DGAVCDec?

Not sure, it might be the extra overhead of feeding it through to a NLE , and avisynth, and the import plugin (Everything is slower in NLE than media players). You can try DGAVCDec, but DSS2 has way higher FPS decoding on multithreaded system like a quadcore

Another "trick" you can do is to use something like MPEG2 as a lossy intermediate, then swap the asset (replace footage back to the original version) in the clip bin after you're done your editing (all instances will be replaced in the timeline, and edits will be kept), so when you export you get full quality. The lossy proxy has to be the same size dimension (can't use 1/2 sized proxies etc...)

[creamyhorror]

Not sure, it might be the extra overhead of feeding it through to a NLE , and avisynth, and the import plugin (Everything is slower in NLE than media players). You can try DGAVCDec, but DSS2 has way higher FPS decoding on multithreaded system like a quadcore

I guess Ut is the way to go, in the end.

Another "trick" you can do is to use something like MPEG2 as a lossy intermediate, then swap the asset (replace footage back to the original version) in the clip bin after you're done your editing (all instances will be replaced in the timeline, and edits will be kept), so when you export you get full quality. The lossy proxy has to be the same size dimension (can't use 1/2 sized proxies etc...)

Thanks, I've been meaning to look into proxy files for a while. This should help a good deal.

[foochuck]

https://forum.videohelp.com/topic346331.html

Or Xvid4PSP

or the easiest (and best, but not free) is NeoScene

I have Neo Scene

If I convert all of my footage from AVCHD to AVI with Neo Scene, can I then resize the AVI to SD AVI files?

[minidv2dvd]

you have to crop or pan/scan the sides off them. just resizing will squish the video to the wrong aspect ratio.

[poisondeathray]

Are we assuming this is for DVD and you want to keep interlacing?

As already mentioned, if you resize without an interlace aware method you will get distortions (It's ok to simple resize only along horizontal axis, but that's not very useful)

If you use Adobe for everything, you will get lower quality results - not because of the MPEG2 encoder - it's because it's resizing and deinterlace-reinterlace algorithm is poor. See here :
http://www.precomposed.com/blog/2009/07/hd-to-sd-dvd-best-methods/

(I can't speak for the hd2sd package or if it's any good, but I've used the other avisynth method I listed above):

[foochuck]

Poison, what software will I need to download to do the method you described below?

In this thread, poisondeathray suggested this Avisynth script: .....

If he's starting from AVCHD, it's already YV12 so ConvertToYV12() isn't necessary (as opposed to DV in the other thread, which was YUY2)

foochuck would probably want to re-interlace it as well, assuming the end goal was NTSC DVD

AssumeTFF() 
interp=NNEDI2(field=3) #double rate, starts with top 
YadifMod(order=1, field=-1, mode=1, edeint=interp) #TFF,field set to order,double rate 
Spline36Resize(720,480) 
SeparateFields()
SelectEvery(4,0,3)
Weave()

Another option is to use the Premiere AVS Import plugin, which works with CS3, and you can get practically any file into it that way, including AVCHD.

But the benefit of resizing to an SD lossless intermediate is that editing will be a snap even on old PC's

[poisondeathray]

Poison, what software will I need to download to do the method you described below?

avisynth and the following plugins: yadifmod, nnedi2 which you can find here http://web.missouri.edu/~kes25c/

It might be easier to use that hd2sd package and guide from that weblink (although I've never tried it out). If I ever do try it out I'll post some comparisons, but I almost never make DVD's anymore.

[foochuck]

I looked at that link - it looked like it was geared towards making MPEG-2 footage - whereas I want to end up with SD Interlaced AVI files (1.2 PAR) that I can edit in Premiere Pro.

[poisondeathray]

But what are you going to do with SD interlaced AVI files? Just play with editing? ie. What is your end goal?

I would use the script to encode to a lossless intermediate (e.g. huffyuv, lagarith or UT is even faster on quads), then import into premiere and conform the file to your PAR

But a warning: that script is very slow because it uses a higher quality filters, that's how you get rid of the artifacts in the first place. If it's too slow for you, you can make some trade offs (e.g. use yadif instead of yadifmod+nnedi)

[foochuck]

I'm going to be editing the SD interlaced AVI files in Premiere Pro. The rest of my footage is in that format. After I'm done editing I will be saving everything to MPEG-2 for DVD format.

[foochuck]

I'm working on a 64-bit system with about 8 GB of ram. I'll try out this method tomorrow. Thanks your input / help.

But what are you going to do with SD interlaced AVI files? Just play with editing? ie. What is your end goal?

I would use the script to encode to a lossless intermediate (e.g. huffyuv, lagarith or UT is even faster on quads), then import into premiere and conform the file to your PAR

But a warning: that script is very slow because it uses a higher quality filters, that's how you get rid of the artifacts in the first place. If it's too slow for you, you can make some trade offs (e.g. use yadif instead of yadifmod+nnedi)

[poisondeathray]

What field order is your other footage? e.g. DV is BFF

You might have a problem with mixed field order with different footage, as AVCHD is TFF, and the script above conserves TFF for SD. I think there is a procedure to field order swap in avisynth, but I am uncertain...

[foochuck]

My other footage is SD DV - Interlaced lower field first.

[poisondeathray]

This seems to work: Get jdl-util.avsi and jdl-interlace.avsi from the url below, put in avisynth plugins directory (just like you did with nnedi2.dll, yadifmod.dll) , and use the JDL_ReverseFieldDominance() function. The 1st and last field will be screwed up unless you pass a bobbed frame JDL_ReverseFieldDominance(bobfirstlast=true)

http://avisynth.org/stickboy/

I tested it out on some avchd and seems to work ok => now lower field first instead of top field first

If you do it first before the deinterlace-resize-reinterlace step, the arguments in yadifmod and nnedi will change to reflect BFF instead of TFF. If you do the field swap after, they will be the same. I'm not sure if one method is preferrable to the other

[foochuck]

The output from the script will be progressive 720x480. You can import that into your editor, edit it, and interlace once again if needed. Follow Baldrick's method of installing VirtualDub and the Lagarith codec (you should find a guide).

Do I need Baldrick's method before or after I edit my video? Is that used to re-interlace my footage???

[poisondeathray]

Do I need Baldrick's method before or after I edit my video? Is that used to re-interlace my footage???

You want to keep it interlaced, but just a HD=>SD proper conversion so you still need to encode it to a lossless AVI intermediate using something e.g. vdub

So install avisynth, plus the plugins mentioned earlier in the plugins directory. I already assume you have ffdshow + haali splitter installed for decoding .mts or .m2ts files, then type the following in a notepad .txt file, save it, then rename the extension from .txt to .avs. Change the filenames and paths to match your files

DirectShowSource("0001.mts")
AssumeTFF() 
interp=NNEDI2(field=3) #double rate, starts with top
YadifMod(order=1, field=-1, mode=1, edeint=interp) #TFF,field set to order,double rate 
Spline36Resize(720,480) 
SeparateFields()
SelectEvery(4,0,3) 
Weave() 

AssumeTFF()
JDL_ReverseFieldDominance(bobfirstlast=true)

Now open that .avs file in virtualdub. Select video=>fast recompress, video=>compression (then select either lagarith, or huffyuv or ut video codec or some other lossless format, you may have to install them if you already haven't. I highly recommend UT video codec for lossless if you have a quadcore, its many times faster than lagarith and almost twice as fast as huffyuv-mt for decoding), then file=>save as avi. If you leave audio as direct stream copy, it will encode the audio as wav. If you wanted to disable this, just go audio=>no audio. Then you could import the native audio separately into premiere (you might demux it using tsmuxer for example)

As I mentioned before, this script is very very slow. If you want a faster version, but slightly lower in quality (when you resize hd=>sd , many of the artifacts get reduced so they are not as visible). You can do some mini-tests to see if it's an acceptable tradeoff for your tastes.

DirectShowSource("0001.mts")
AssumeTFF() 
Yadif(order=1,mode=1)
LanczosResize(720,480) 
SeparateFields()
SelectEvery(4,0,3) 
Weave() 

AssumeTFF()
JDL_ReverseFieldDominance(bobfirstlast=true)

When you import into premiere, you have to conform the clip to the proper PAR (1.21), field order (lower field) etc...by using the "interpret footage as" function, because it will initially be interpreted as progressive and square pixel.

Cheers

[foochuck]

I hate to be a PITA (pain in the a$$) but what do I feed this Avisynth script into? Virtual Dub? Premiere Pro?

If you're starting out at interlaced 1080 and you want max quality, you should bob-deinterlace before you resize. (Resizing without deinterlacing damages field structure.)

In this thread, poisondeathray suggested this Avisynth script:

AssumeTFF()
ConvertToYV12(interlaced=true)
interp=nnedi2(field=3) #double rate, starts with top
yadifmod(order=1, field=-1, mode=1, edeint=interp) #TFF,field set to order,double rate
Spline36Resize(720,480)

This is very good deinterlacer, but also very slow. You'll have to install the Yadifmod and NNEDI2 plugins from http://bengal.missouri.edu/~kes25c/.

The output from the script will be progressive 720x480. You can import that into your editor, edit it, and interlace once again if needed. Follow Baldrick's method of installing VirtualDub and the Lagarith codec (you should find a guide).

edit: If you use soopafresh's first method, note that you'll get 848x480 square pixels rather than 720x480 with PAR 1.2. Or you can choose the 1920x1080 output option, which means you retain the full resolution for importing (higher quality but much more taxing for editing).

[poisondeathray]

I hate to be a PITA (pain in the a$$) but what do I feed this Avisynth script into? Virtual Dub? Premiere Pro?

You might have missed it - I gave detailed instructions right before your post ^ . avs script into vdub=> spits out a lossless sd-avi with lower field first => import that avi into premiere with the same sequence settings as your other dv footage

Although you could feed the .avs script directly into Premiere with the avs import plugin, you wouldn't get real time editing (i.e too slow to process). So in this case, it would be better to encode to a lossless AVI intermediate for import

[videobread]

Just my 2 cents but, this thread is so complex it's almost comical. I had to ask myself why shoot in HD at all when the end product is SD. I sell DVD's - that's SDVD. AVCHD and HDV are shortcut manufacturer sponsored formats. Editing software is not yet prime time. Rendering/hardware costs are out of sight. Blu-Ray blanks are not mass marketable, not printable for the money and the cost per unit is prohibitive. Maybe you should consider shooting in SD and waiting for Adobe to develop their product more, for hardware prices to come down and for a market for Blu-Ray to develop. Is this a commercial venture? Are you trying to make a buck? Sounds like you are spending a lot of time and money to go the long way around the barn.

[foochuck]

Videobread: I shoot everything in HDV. For this project I had a camera problem and was forced to shoot with an AVCHD camcorder. I have a normal workflow with my main camera for shooting in HDV and from the start I know if I need to edit in HDV or SD. I have workflows for both.

It's my first time using AVCHD and I'm trying to figure out a workflow for this format.

[foochuck]

I hate to be a PITA (pain in the a$$) but what do I feed this Avisynth script into? Virtual Dub? Premiere Pro?

You might have missed it - I gave detailed instructions right before your post ^ . avs script into vdub=> spits out a lossless sd-avi with lower field first => import that avi into premiere with the same sequence settings as your other dv footage

Found it! Thank you sir.