Need some help with a tricky time-lapse

U319

TPF Noob!
Joined
Nov 15, 2012
Messages
4
Reaction score
0
I am new to time-lapse and have a challenging problem I could use some help with. Please let me list my equipment: I have a Watec WAT-250D2 CCD camera, a 28 mm lens adapter, and a converter for composite/S-video to USB 2.0. I have GrabBee software.

I need to do the time-lapse over several days, however there are rapid movements that I want to try to capture within this long time-period. I thought, therefore, that a constant video approach would be best, speeding up the video later. Here are some problems I'm running into: In preliminary tests, I'm finding that file size increases 16GB/hour of video. Over 7 days, this is almost 3TB. The time-lapse needs to be high quality (or as good as I can get). My software allows me to save in AVI, MPEG-1 and 2, VCD-PAL, SVCD-PAL, DVD-PAL, WMV, and MP4. The video size and bitrate available changes depending on file type, and it seems MPEG-2 is best at 640X480, 9000 kbps, and 29 fps. I noticed that if I decrease the bitrate, I save a lot on file size, but are there any other ways to reduce the file size? I think I want to keep frame rate at 29 because for the time lapse to be a reasonable amount of time post-editing, I think it will have to be sped up a huge amount, and it will be super fast. If I'm starting with something shot at 12 fps, post-editing I'll probably see nothing/it'll be too fast, right?

Is it possible to shoot video in intervals? For example, maybe 5 seconds of video every 30 seconds? I imagine this would not edit together very smooth, and is it even possible to edit them together?

Thanks for your time.
 
You're not going to be able to distinguish rapid movements when you speed it up a lot... Why do you need to capture these movements?
5 seconds every 30 seconds is going to give a rather weird movement pattern when put after each other and sped up, the movie won't appear fluid because of constant speed variations.

I don't really see why 12 fps would be much of a problem here...
If you speed up 12 frames per second you'd only need to triple the speed to get a fluid movie (which still is over 2 days long, given the original footage is 7 days of shooting).
Say the human eye can detect movements of about 1/60 a second (rough guess, might be too small to actually detect but a lot of screens are 50 or 60 Hz so more is kind of useless anyway) and say we shoot for 7 days at 12 fps.
That would give us a total of 7 x 24 x 3600 = 604,800 seconds, which results in 7,257,600 frames.
Now say you want to make a movie of about 3 hours long (that's usually the longest a serious movie gets, with of course the odd exception).
3 hours is 10,800 seconds. Spread the frames over those seconds and you get 672 fps.
That's 56 seconds of footage per second of movie. So, any movement lasting about a minute in real life will last about a second in your movie.
Say we're using a screen at 60 Hz that means one single frame will cover one second of footage. This means that movement lasting a second in real life will not even be movement at all in the movie!

What exactly are you trying to achieve with this project again...?
 
Thanks very much! Thanks for the math. I understand I'll lose the rapid movements as they look in real-time, but I thought at 12 fps they would be lost completely (i.e. not have an effect on the final movie) after being sped up. For example, I am watching something grow. It takes about a week to reach maximum size, and in between it sort of increases and decreases rapidly (as it gets bigger). At 12 fps I'm thinking that you will just see an expansion. And at 29 fps, I hoped maybe you would at least see a sort of flowing increase-decrease up to the maximum size (?).
 
Also, from your math, I don't think it's reasonable for me to shoot video for 7 days, when I want the final film to be under 30 mins. To really see everything, now I'm imagining doing 1 timelapse where I film (or take still pictures at intervals) to capture the whole growith cycle; and then overlay other timelapses on top of this -- that run at slower rate -- to show the expansion at different time points in the growth cycle. Not sure if that's clear. This is for a science project.

Maybe I could shoot the whole thing in constant video -- speed it up a million % for the 30 min video, and then do half-that% or something for the overlays. This way I would only have to edit 1 file.

If I did stills for the big one though, would I have better quality? Probably not with my CCD camera, but I don't know.

Thanks.
 
I say shoot at 4fps or somewhere around there. Unless you're trying to watch a time lapse of your girlfriends house for say, a week of you being on vacation, you'll easily be able to see tiny amounts of movement. If I shoot a twenty round burst with my camera at around 4fps, that's five seconds worth of time. I can promise there will be not one single change in those 5 seconds worth of images. Hell, I seriously might do this later just for verification.
 
I'm guessing that by flowing increase-decrease you mean that in your movie it will appear to grow fast before your eyes without the images visibly jumping to other sizes.
This is indeed controlled by the fps. Not the shooting fps however but the final movie fps.

To get fluid motion in a movie you'll need to get a high enough fps output so that the human brain will see it as fluid motion.
The bare minimum here is around 30 fps I think. You're talking about half an hour of finished movie so we can calculate how many frames that is.
30 x 60 = 1800 seconds.
1800 x 30 = 54,000 frames in total.
This means that you'll need to divide the entire growth process over 54,000 frames to make it look like motion.
The growth process takes 7 days which is 604,800 seconds (see earlier post).
54,000 frames divided over
604,800 seconds is 604,800 / 54,000 = 11.2 seconds per frame (note: this is the exact opposite of frames per second).

So, if you shoot one photo every 11 seconds for 7 days and put them all together in a time-lapse you'll get a completely fluid and flowing movie of about half an hour (give or take a few seconds).
 
I'm guessing that by flowing increase-decrease you mean that in your movie it will appear to grow fast before your eyes without the images visibly jumping to other sizes.
This is indeed controlled by the fps. Not the shooting fps however but the final movie fps.

To get fluid motion in a movie you'll need to get a high enough fps output so that the human brain will see it as fluid motion.
The bare minimum here is around 30 fps I think. You're talking about half an hour of finished movie so we can calculate how many frames that is.
30 x 60 = 1800 seconds.
1800 x 30 = 54,000 frames in total.
This means that you'll need to divide the entire growth process over 54,000 frames to make it look like motion.
The growth process takes 7 days which is 604,800 seconds (see earlier post).
54,000 frames divided over
604,800 seconds is 604,800 / 54,000 = 11.2 seconds per frame (note: this is the exact opposite of frames per second).

So, if you shoot one photo every 11 seconds for 7 days and put them all together in a time-lapse you'll get a completely fluid and flowing movie of about half an hour (give or take a few seconds).

Just wanted to thank eveybody for their replies and give an update on progress and also discuss some new problems. I learned a lot of information really fast thanks to you guys, so thanks a lot. The goals for the project changed somewhat and now I'm only shooting 4 days -- and I'm shooting in time-lapse mode with my video capture software 1 picture every 2 seconds. The camera is set to 29 FPS. I optomized my computer for time lapse -- turned off sleep mode, etc. I also have it connected to an external hard drive. But the software doesn't seem to be able to make it through 10 hours of shooting without crashing (I have to schedule daily 10 hour shoots because that's the maximum the software allows). I think the software is garbage, maybe. Can somebody recommend good software for this? Thanks.
 
Hi,

Could you post more details about your setup ?

I don't know much about Watec cameras, but some like Canon P&S tend to crash after a long time of continuous operation. It's not really related to the software but caused by the camera itself.

It could be a similar issue.

One option is to power-cycle the camera as soon as capture stop working (it you automate this you don't even need to do this manually).

But I don't know if this could be an option in your situation.

F.
 

Most reactions

Back
Top