This content has been marked as final. Show 177 replies
I think our first specific task should be to gather some information on the various current work flows now used by people interested in the dv2Film process. For example, my own process is to edit in Premiere, then grade in AE, then Author in Encore. The dv2Film process needs to be added in there somewhere.
So, what work flow are others currently using, and how do you see dv2Film fitting into the process?
I have the first results of some codec tests.
The original clip used was a DV export from Premiere. This was run though the dv2Film process using various output codecs and frame rates, including DV at 24p, DV at 30i, Lags using RGB at 24p and 30i, Lags using YUY2 at 24p and 30i, and Lags using YV12 at 24p and 30i. Eight separate runs in all.
The output of each process was brought into a matching Premiere project and exported through the AME at CBR 7 to eliminate transcoding variables. Each of those outputs was then brought into an Encore project and authored to DVD, where the clips were viewed on a properly calibrated 27" CRT TV coming from a standard DVD player (no progressive).
I did not pay attention to the order of the clips, and although it may not qualify as a "blind study" in scientific circles, I was not aware of which transcode was which when viewing. I listed my notes for each version, and then went back to compare.
The result rather surprised me. For starters, there was very, very little difference between the clips. So little that if someone were to jumble them up and ask me to watch them again, I can't say I'd write the same notes for the same clips.
The good news here is that DV may not be that bad a choice for output. It renders faster than Lagarith, which is nice. The surprising news is that for two of the clips, I made the note that the motion looked just a little bit smoother, and on a couple of others, I made the note that the motion had just a hair more stutter. When coming back to compare my notes with the clips, I found that the two I noted as being smoother were both 30i transcodes, and the ones I noted as having slightly more stutter were 24p.
I'm actually a little disappointed in those results. Grading a 24p clip in AE goes much faster than grading a 30i clip, but now it appears that 30i output from dv2Film (do32=true) may be the preferable method.
I'm very curious now to see these same clips on a calibrated, 36" CRT TV from a progressive player.
> output from dv2Film (do32=true) may be the preferable method.
No way. If there is any more "stutter" on the 24p clips it is only because of your DVD player and/or your TV. Even if this stutter is real, you yourself said that it is extremely hard to tell the difference on your interlaced CRT. Why incur the extra time -- not to mention degraded quality on progressive displays -- for something you can barely notice, if at all? Doesn't make sense to me.
Hollywood movies are encoded at 24p. If it's good enough for them, then it's good enough for us.
> Lags using YV12
Don't use that. Use YUY2 or RGB24.
> DV may not be that bad a choice for output. It renders faster than Lagarith, which is nice.
How much faster? Is mulithreading enabled in the LAGS settings? Did you try using DebugMode and frameserving to VirtualDub as YUY2 (+ "Fast recompress" in VDub) instead of compressing directly to LAGS from PPro? I bet it's faster.
You could also try HUFFYUV -- bigger files than LAGS, but reads and writes faster.
My only real worry about DV in your case is that you're planning to export once from PPro, then again from AE. I'd look for a way to skip these additional two DV generations if possible.
>you yourself said that it is extremely hard to tell the difference...Doesn't make sense to me.
Which is why I am processing the current wedding as true 24p. The severe extra time involved with grading a 30i project isn't worth the what little difference I may have seen, and that with a direct comparison, which the bride will not have.
>Don't use that. Use YUY2 or RGB24.
I'm curious as to why. The one note I had for the RGB file is that it may have looked a little dark compared to the others. Both YUY2 and YV12 processed in about the same time, produced smaller files than RGB, and showed no visual difference from the original that I could notice. Those same "benefits" hold for DV as well, by the way, with even smaller files sizes and faster processing time to boot.
>How much faster?
Coming out of Premiere with DV source, 8 to 10 times faster. Going through the dv2Film process, about the same.
>Is mulithreading enabled in the LAGS settings?
Yes, but it didn't seem to take advantage of it. Processing dv2Film the standard way seems to be SMP unaware. I did not try DebugMode out of Premiere yet.
>I'd look for a way to skip these additional two DV generations if possible.
That is a real concern, and more testing will be done.
Is there anything special that I need to do for widescreen footage? I imported widscreen footage and it came in as 4:3. Thanks.
4:3 squished, or 4:3 letterboxed?
> I imported widscreen footage and it came in as 4:3
try intrerpret footage, pixel aspect ratio ?
Have you sent it through to authoring yet? Your DVD player should unsquish it for you, no?
> ... 8 to 10 times faster. Going through the dv2Film process, about the same.
Huh? I just tried a few tests compressing with LAGS in YUY2, and compressing with various DV codecs from VirtualDub. Lagarith did not take especially longer than any of the DV codecs.
> Coming out of Premiere with DV source...
I tried exporting in Premiere from DV -> LAGS, LAGS -> DV, DV -> MPEG2, LAGS -> MPEG2 ... Lagarith was only about 40% slower, maximum, at any time.
I'm not exactly sure what is going on with your results, Jim ??
Hi Jim & Dan
I came in a bit late on this thread and just got up to speed with your previous (part I) trials. One of the most interesting and
informative threads I've seen.
I haven't done any trials yet, but will attempt to this evening. However, my knowledge of scripts, plugins and filters is very limited.
I am going to spend today trying to learn more on it.
Meanwhile, can you post a before / after clips illustrating your results?
>Hollywood movies are encoded at 24p. If it's good enough for them, then it's good enough for us
But Hollywood starts with a 24p source. We're starting with a 30i source and then processing it to create 24p. We're using several different tools to do that, each with an array of possible settings, and we essentially just started doing this.
Hollywood uses highly-trained, highly-paid and experienced engineers who do nothing except telecine film for DVD. They have very expensive encoders that make ten or more VBR passes over the material when transcoding for DVD. We have encoders that can't make more than 2 passes, and which were designed to encode video, not film.
Hollywood has cinematographers that know how to shoot film in the first place, and they've been doing it for a very long time. We have...us. :) And despite Jim's education and wedding production experience, I doubt that he shoots video as well as a studio's cinematographer shoots film.
So our results may be skewed from the beginning because of the inherent limitations in our workflow. I suspect it will take a while to get this down to solid, reproducible results for everyone.
> But Hollywood starts with a 24p source. We're starting with a 30i source
Yes, yes, of course... :) I only meant that our DVD players are routinely called on to add pulldown and we (the viewing public) don't notice or care.
> So our results may be skewed from the beginning because of the inherent limitations in our workflow
Right again. I never once said that the results will be perfect or look "just like film" or anything of the sort. I'm just trying to get the best framerate conversion I can, given the source.
>Lagarith did not take especially longer than any of the DV codecs.
Coming out of VirtualDub, I agree. The longer time is coming out of Premiere originally. Using DV media, a DV export is much faster than a Lags export.
>can you post a before / after clips illustrating your results?
Working to have them up today.
>I doubt that he shoots video as well as a studio's cinematographer shoots film.
I heartily concur.
>I suspect it will take a while to get this down to solid, reproducible results for everyone.
Hence this thread. The more testing people do under various conditions, the more the process can be developed into a smooth, workable experience for everyone.
> The longer time is coming out of Premiere originally. Using DV media, a DV export is much faster than a Lags export.
Right. But if you're converting a finished 60i timeline... that's why you should use DebugMode -> AviSynth -> VDub (LAGS) and do the 24p conversion right there. Save yourself a step and an intermediary compression stage.
That is definitely a test on my list.
I'd still like to get some answers to question number one - Malick, KMS, tlc?
What kind of work flow do you now employ, and where do you imagine the dv2Film process might best fit in?
My workflow is shoot everything 60i and edit everything in 60i. Once every step is complete... editing, color correcting, audio, etc...(remembering to make all my cuts on frames ending in either 0 or 5 so I don't get unintentional cross dissolves) I export a Microsoft DV AVI and then process that through DVFilm maker twice. Once for 24p and then once to make it 60i again. I like using DVFilm maker for this because I like the look of 2:3:3:2 pulldown over 3:2 pulldown. I then import that file into premiere and then export to MPEG2-DVD so Encore will not have to transcode anything.
I find this a lot faster then trying to convert every clip to 24p to edit with. I rather do it with one file at the end.
> because I like the look of 2:3:3:2 pulldown over 3:2 pulldown
That's interesting. Generally, 3:2 is associated with a better "look", while 2:3:3:2 is used for ease of editing.
Anyway, 2:3:3:2 can easily be accomplished with my script also:
SeparateFields().SelectEvery(8, 0,1, 2,3,2, 5,4,5, 6,7)
In the next version I'll have a "built-in" 2:3:3:2 option. (say, "do2332" ???)
>I like the look of 2:3:3:2 pulldown over 3:2 pulldown.
That is contrary to conventional viewpoints. Still, thank you for the post.
>that's why you should use DebugMode -> AviSynth -> VDub (LAGS)
Ok, I'm afraid I'm gonna have to play the idiot here Dan. How, exactly, do I do this? I've done the following:
2. Select DebudMode FrameServer under File Type
3. Choose YUY2 and Write Audio as options.
4. Hit Next
5. Signpost file is created
6. Now what? I tried pointing the .avs script to the signpost file, but that would not open in Vdub. And I don't understand enough to know why, or how to fix it.
Sounds correct to me. I do it all the time. What's the error you're getting?
I have only been playing around with some test clips right now, but so far my favorite method is:
1) Capture 60i footage
2) Use AviSynth Batch Scripter to create a .avs file for each clip
3) Import .avs files into a 24p project (edit and color correct)
4) Export 24p DVD to Encore
Jim, what graphics card to you have? Just wondering...Looks has been solid for me PPRO.
My problem may lie elsewhere. Vdub will now not open the .avs script no matter what file I point it to, including those that previously worked. Could the Premiere AviSynth plug-in or DebugMode be interfering somehow? Those installs are the only things I can think of that changed.
It may also be relevant that Premiere would not open the .avs file, stating that I needed a newer version of AviSynth. However, the newest 2.5.7 is the one installed.
I'm using an X1650Pro card.
> Could the Premiere AviSynth plug-in or DebugMode be interfering somehow?
I can't imagine how. I use them both all the time.
Try a very simple script like this:
..or like this:
Do either work in VDub? If not, is there an error message?
Likely your AVS plugins folder is smurfed. I had the same problem the other day. Make a backup copy of it, then delete it and repopulate it with Dan's packaged plugins.
"foo" don't work. The error is the same regardless of what the .avs has typed in it:
Avisynth open failure:
Avisynth: script open failed!
I notice that a simple line like Version is supposed to create a 10 second media clip playable by Windows Media Player showing what version of AviSynth is installed, but even that doesn't work.
Something got fubarded.
I reinstalled Avisynth and the newest version of your script with plug-ins. I am now getting a different error (which might actually be progress.)
"Could not open video stream in any supported format." And it pointed to line 35 of your script as the culprit.
> And it pointed to line 35 of your script
That's the line (sure enough) where it opens the .AVI file. Does this only happen now with the frameserver output? Are your regular DV files working again?
I didn't try with FrameServer yet. It happens with DV files that once worked fine.
Reboot? Uninstall DebugMode, Premiere Plugin, AviSynth -- reinstall?
Can you open the DV avi files in VirtualDub -- non-DV avi files?
I have NEVER had that kind of problem, and I've been using all of these for years. Yes, something got smurfed somehow.
I can try to help you figure it out if you like, but it's not efficient here. Write me @ d.isaacs --at-- comcast --dot-- net if you need more help on getting things working.
>@ d.isaacs --at-- comcast --dot-- net
Using the PPRO plugin for Avisynth is out for me. Too slow and buggy. It made Premiere crash and totally disapear...
I wish there was a way to get 16x9 to come out correct. It's a pain to change each one manually in PPRO. Any ideas?
Does scaling the video on the timeline have more of an impact on the quality of the video than using the Interpret Footage dialog box? If not, I can just scale one clip and paste attributes to the rest of the clips on the timeline...
> It's a pain to change each one manually
You can select multiple clips in their project window and change their aspect ratio all at once.
> Does scaling the video on the timeline have more of an impact on the quality of the video than using the Interpret Footage dialog box
Yes. It works a totally different way when it comes to exporting.
So, KMS... Just modify your earlier workflow a tiny bit:
1) Capture 60i footage
2) Use AviSynth Batch Scripter to create a .avs file for each clip
3) Batch process the .avs files in VirtulDub to DV
4) Import DV files into a 24p project (edit and color correct)
5) Export 24p DVD to Encore
Also... about your 16x9 flag thing. There is a program called DV Date Time Changer that can change date/time and flag the aspect ratio of DV .avi files. It works, but there's no batch function.
You can check DVMP Pro 2 out also, but it is not free.
Sorry - I have a deadline looming.
Jim - looks like my trials will happen over the weekend.
Will report back on Monday. Hope you plugins folder is "unsmurfed"
Couple of questions:
Dan, I am in PAL-land so my conversion would be from 25fps to 24p right?
Can Dynamic link not be used to simplify the workflow?
Also where does Audition / Soundbooth fit in?
I usually do sound sweetening right before exporting to Encore.
> Dan, I am in PAL-land so my conversion would be from 25fps to 24p right?
Oh.. this won't work for you then (not yet). It can output PAL Film 25p, but it's meant for NTSC sources.
PAL 50i -> 25p or 24p is really much simpler. In the first case, you simply deinterlace the video. Then to get NTSC 24p output, you simply slow that 25p down by 4% (95.904095% of the original)
I'll put a dv50i2film() function in the next version. Very easy to write.