@duartix Thanks for the quick answer! I feel much better now. :)

@zcream: Thanks for pointing it.

Yes it works, even though it might not be the most efficient way. Let me elaborate:

Given the circumstances (light and shutter speed) he/she can't use a 360ºshutter (unless using some ND filters). There is just so much light. So, timebuster (the 24p version) might be more adequate and easier on the battery as there are less frames to process and no danger of interlacing problems later in post.

If he/she knew in advance that in the final result there was going to be such an extreme time compression/frame dropping (50:1) he/she could also have changed timeHBusteR to use a much longer GOP and that would have saved a lot of bitrate. In the recording there is 1/5 good frames, in the output it drops to 1/50 (according to the script). No point in saving frames you are dropping later in post, unless you are frame blending, which I believe wasn't the case by looking at the script. But sometimes you just don't know in advance... ;)

For 1/50th shutter and 25P, what GOP would you recommend ? I am using this on a tripod to film traffic and ocean timelapse. So there should be a lot more temporal continuity..

@zcream It depends! I'm assuming from your shutter speed (SS) that temporal/motion continuity is an intended casualty. If you are using a 1/50s SS at 25fps, you won't be recording half the time.

This is where I should advise you not to do it unless it's the way you intend it to be... And why? Because if you are filming traffic, your light trails will look something like this: _ _ _ . It will look choppy (like animation choppy).

If you are into motion continuity, then you should go with timeHBusteR and choose a SS speed that gives you continuous motion (check those PNG files @lr1k mentioned). Next you should set the GOP to marry that SS. After capture there is another requisite to get fluidity though, when you are accelerating beyond the GOP scale (ex: if you choose 25p at 1/2.5s SS which according to the table should use GOP 10, and you go beyond 10x acceleration), then you must do frame blending (instead of frame dropping) to get continuity. The principle is simple: to get continuity, all motion must be captured (100% on those png GOP tables) and all frames with unique content must contribute to the output.

Imagine the following content at 5fps where = denotes a trashed frame with the same content as the preceding letter, only with very low quality (this is the way timebuster works to save bitrate)

A====B====C====D====E====F====

The simple post processing is accelerating 5x with frame dropping because you want to get rid of those low IQ = frames, and will get you this:

ABCDEF

Accelerating to 5x assures motion continuity because frames A,B,C,D,E,F are full time capture and full quality while the = are trash that is there just to satisfy the codec, which needs to encode 25fps.

Accelerating to 10x can't be done in the same step (at least in After Effects) because if you frame drop you'll get:

ACE (no continuity)

And if you frame blend you'll get:

(A====B====) (C====D====) (E====F====) (low quality frames included in output)

In After Effects you need to frame drop first to get

ABCDEF

and then frame blend it to get:

(AB)(CD)(EF)

If you are using After Effects check out the workflow within this thread because it's not completely trivial to weed out the bad frames as they might be very similar to the good ones.

OTOH if you don't care about motion continuity and you have an idea of the final acceleration (ex: 15x), you just might use GOP15 saving a lot of bitrate and then in post you just choose 1 out of every 15 frames (but making sure you are selecting the I frames and weeding out the bad ones with frame dropping). This is the trivial workflow. :)

A small clip with clouds passing by 2.02 used, speeded up to 3000% in premiere pro Slightly graded. Used ND 106 + POL iso 160, F4.5 1/2.5 25mm 0.95 lens

Cheers @vicx.

If you are using timeHBusteR 2.02 at 1/2.5s you needn't change "1080i60 GOP Size" as it's already at 12. If you changed it from 10 to 12 then maybe you mistook timebuster for timHBusteR and you are using the 24p version when what you want is timeHBusteR @ 30fps (HBR) NTSC.

Let's get this one out of the way before we look into your I-Frame size (which I can't really explain so far).

If that is not the reason, please upload your changed settings so I can have a look.

Thanks for the report.

@tida re: AvsPmod script error(s). Running Windows 7 (64 bit) on Intel Ivy Bridge i7.

I loaded all the specified apps - AviScript (32 bit 2.5.8.5) and included the 64 bit mods, AvsPmod 2.3.1.0 x86, MeGUI_2153_x86. I then loaded the Timebuster 24p 2-5s v10 script into AVSPmod (see the attached screen shot) and followed your screen shots...

No mater what I do I get the following AviSynth error (see the screen shot) which claims it was unable to locate the proper filters when loading the settings into MeGUI. I can transcode MTS files with MeGUI, extract audio and such so MeGUI seems to work.

It's looking like something to do with "DirectShowSource" - There is a lot of chit-chat in the appropriate forms concerning if/how to make DirectShowSource work with Win 7 64 bit with conflicting claims...

Question: Were you able to get your AviScript solution to work with Windows 7 (64 bit)? If so would you provide relevant details? Thanks in either case :-)

@vix Thanks! I got it running with the x86 versions and the Timebuster 24p 2-5s v10 script. Figured it had to be a 'CoDec' and was gonna go the Haali/ffmpeg route. What gave me pause to continue is I already had the LAV filters installed however they were local to MPC-HC. Installing them system wide did the trick. The reference to program files rather than program files (x86) was a hitch (win 7 64 bit) but edited the AVS script to reflect that. Being new to the AvsPmod/Avisynth/MeGUI workflow caused confusion on details that is obvious to the seasoned user. It certainly wasn't intuitive to the 'casual user'...

Correction: Assumed I had it running properly :-( The speed-up works fine but the overall file time is the same as the original. It appears that the MP4 output file encodes with the proper 'step' parameters but continues to run from the last frame until the original file time length is reached. So ya get the time compressed video until the last frame and then a still frame until the source MTS file length is obtained. I gotta be missing some parameter somewhere... Any ideas?

NEVER MIND - Killed the Audio Muxing and all is fine :-)

I hate to keep talking to myself but... Finally got TB 2.02.1 to merge with Valkyrie 444 TYPE-ZERO2_Candidate1.5. The issue was with the settings in the 24h TimeHBusteR 2.0 setb.ini file. I had assumed that the changes to TB 2.02.1 alpha had been incorporated in the setb.ini (24h timeHBusteR Base 2.0) of 24h TimeHBusteR 2.0.zip located in the Hack Vault. Not so :-(

I have updated the base 24h timeHBusteR 2.0 setb.ini to reflect proper operation when merged with Valkyrie 444 TYPE-ZERO2_Candidate1.5 (that is all I've tested so far). 30HBR time lapses OK while 24P Cinema, VMM 80% work as normal modes. See the attached updated setb.ini which include the TB 24h timeHBusteR 2.0 settings extracted from the merged 24h TimeHBusteR 2.0 setb.ini.

Confused? What I said is the attached setb.ini will merge with Valkyrie 444 hacks without wonking the 24P settings.

I am still using FCS3 - FCP7. I used timeHBusteR out and 1/5s at PAL 25p In Timeline1 I set the clip at speed of 500% with no frame blending. Problem is that FCP defaults to 499.76 or something random. Any idea how I can force exactly 500% ? My files are transcoded to Cineform. altenatively is there a better workflow in speeding up AVCHD directly ?

@duartix I have your settings in AE. How do you fix the interlaced artifacts ? Frame 0 and Frame 1 were soft, Frame 2 was sharp. So I started with Frame 2. TimeHBuster PAL 1/5s default values. 20% Time Stretch. Do you de-interlace the footage created by AE ? The rendered footage has interlaced artifacts..

@zcream :Follow the workflow for weeding out bad frames: http://www.personal-view.com/talks/discussion/comment/49325#Comment_49325 but before you do that you have to import the footage properly. Yesterday I had that very same interlacing problem and couldn't remember how to do it, but fortunately the internet forgets nothing: http://www.personal-view.com/talks/discussion/comment/57925#Comment_57925