I've never tried this myself, but...

Don't a lot of NLE's do fast motion by simply dropping frames? In other words, if you specify 200% in the speed settings, doesn't the software just display every other frame at the proper 30fps?

I wonder if you shot your half second intervals (15 frames) - once every minute - then played the resulting clip back at 15 times normal - would it look the same as shooting a single frame every 2 seconds?

And if my math is right (It's late after a LONG shooting day!) extending the capture interval from once a minute - to once every 10 minutes - would net you a single frame representing 20 seconds of real time?

Don't rely on my late night math, but there's a concept here tha MIGHT be worth persuing. Or not!

(Tho I agree everyone I know who does time lapse uses digital stills)

I once hooked up a webcam to the Windows Media Encoder and just pressed record. Then I saved that file as a wmv and imported it into Premiere Pro.
Can't remember the speedup, but I think it was 10,000 after the first render and again 2000 or 3000 after a second time.

I recorded about 4-5 hours at 320x240 continuously.

Gonna look it up and if I find it, I'll post it on youtube.
Here it is: http://www.youtube.com/watch?v=dRy-0DhHmYA

Eh...I disagree. One frame at a time would just be letting the camera roll, then speed that up. That's the best way for video to do it. When you start getting really temporal with this stuff like stretching it or squashing over mutiple frames, the more concurrent frames you have the smoother the motion will appear. So, for 2 seconds, or whatever the final result, you get the net effect of simply speeding up a real time capture.

I suppose we could be talking theoretical, but my real world experience with video time lapse captures (limited as it is) says the longer you capture real world motion, the better...

What I'm suggesting will have a different look to it, for sure than .5 every 30 sec. But the camcorders I've used all have to wait 30 seconds between capture - which actually loses a LOT of info.

Nice and smoooooth.
 

And thus the reason it's motion is soo smooth. It just looks like you sped up a day - and that's what time lapse should look like. .5 every 30 sec. has a harder time doing that.

When doing a time-lapse segment with a DV camera I use BTV Pro,
and more times than not capture 1 frame a second.

Here's a quick time lapse car journey I made about an hour ago:

http://www.enosoft.net/video/QuickDrive.wmv

(I was in a hurry so didn't bother the clean the windshield nor put a polarizing filter on....)

I hooked up my PDX10 to a T40 laptop running our Enosoft DV Processor and captured directly to disk at a rate of 1 frame in every 5. Windows Movie Maker created the .wmv file at 320x240x30fps.

The journey comes to a rather abrupt stop when I pull over to talk to the road crew foreman - turns out it is TimeWarner putting in cable - but short of my house by about 1/8 mile. :-(

Black-hole sun wont you come and wash away the rain.....

I couldn't resist rummaging through my drawer full of old Video 8 PAL recordings to find this:

http://www.enosoft.net/video/Zanzibar%20Sunrise%202.wmv

It was recorded on the east coast of Zanzibar in May 1995. I was the only person in our group of 20+ people who bothered to get up early enough to witness the amazing sunrises over the Indian Ocean....

The clip is at x25 normal speed.

John, that is a pretty amazing sunrise...

Anyway, about using the .5 second every 2 seconds for time lapse... It would seem to me that you would just be saving tape. See if you follow my logic or if you think it is bad logic. The camera records 15 frames out of a possible 60 frames over a given 2 second period. If in your NLE you speed the clip up to 15x normal play, doesn't the NLE just drop out 14 of 15 frames? Wouldn't that give you 1 frame for 2 seconds? I think that would work for time lapse, right? I'll dig around and see what it looks like, I've got an old tape from last summer of some people at an event that was recorded like this...

Does anyone watch (or has watched, I don't know if it's on anymore) American Hotrod? They would often do this to show the passing of time. I remember the shot being framed so that the clock was in the left of the frame and the action was in the right. The frames, if you watched the second hand on the clock, were taken at thirty second intervals. Now I know the show has a big enough production budget to have captured time lapse any way that they please, but it would seem to me that speeding up interval footage shot with a cam would be the easiest.

Also, I just thought of this... Has anyone with an XL2 watched the "Fusion DVD: The Ultimate Guide to the Canon XL2?" In this video tutorial of the XL2 the host tells of using the interval recording to make time lapse and it is spot on. So I would think that this technique is viable. I'll make a test and see about posting it.

yeah but of course the .5 seconds was referring to a completely different quantity than the 4 hours. the latter is the total recording time. they are not comparable.

in one example you record 4 hours and speed it up by a factor of x - which is equivalent to recording ONE frame every x frames. very smooth.

in the other you record for 0.5 seconds (i.e 12.5 frames), wait x-12.5 frames (i.e. x is always bigger or equal to 12.5) and then record 0.5 seconds again and do the whole thing for y amount of time. which means you get smooth bits with jumps in between. it's equivalent* to playing back the time lapse in the first example with a frame rate of 2 fps - definitely not smooth!

*assuming that all the frames recorded in the 0.5 second interval are essentially identical

Yeah, I was really surprised to see that. At the time I had a Sony handycam lying around too, but I was a bit scared I might fry the CCD. Anyone has advice on that, about protecting your camera if you still want to capture that sunrise or sunset?

And, to finish off, this one I found on Youtube. I think it's stunning.
http://www.youtube.com/watch?v=UrGcd6PN7EE&NR=1

From what I've read here if your eyes can stand it, so can the camera. For the parts where the lens would be directly in the sun's path, avoid or filter. I point my camera so it is never in the direct path of the sunrise and use what filters I have (UV). The sun is always to the left or right when I film. While it may just be foolhardy on my part, it hasn't seeemed to fry my camera at all...

For the YouTube video, I agree -stunning! That is what I'm shooting for, do those streaks of the car lights come with speeding the video up, slower shutter, different timing or combo -what?

Definitely a slow shutter, but also clearly NOT shot with even a couple seconds between each shot. Lots of slow shutter still pics taken very rapidly back to back would be my guess. The 20D can take something like 5 frames per second, and maybe that's how it was done.

Looks awesome though, no matter how it was done....

I think the titles indicate that the realtime duration was about 2h58. And yeah, I'm on the slowershuter page too. Can this be done with an XHA1?

Your logic is good :-)

The key to getting good time lapse results with a camcorder that records in chunks of, say, 0.5 seconds every n seconds is to ensure that the video is sped up by a factor of at least 15 (for NTSC) or 12.5 (for PAL). That way, only one frame out of the short real-time sections will be used. If you speed up by only, say, a factor of 5, you will end up with a clip where the time between each frame will be vary between a shorter and longer value.

In the 0.5 seconds every 2 seconds case, the frames that get recorded are (in timecode terms - SS:FF):

00:00 *
00:01
00:02
00:03
00:04
00:05 *
00:06
00:07
00:08
00:09
00:10 *
00:11
00:12
00:13
00:14
02:00 *
02:01
02:02
02:03
02:04
02:05 *
02:06
02:07
02:08
02:09
02:10 *
02:11
02:12
02:13
02:14

The *'s are the frames that would be sampled for a speed-up factor of x5. You end up with the following frames:

00:00, 00:05, 00:10, 02:00, 02:05, 02:10, etc

Hence, a time lapse clip created this way would appear jerky. If the camcorder records, say, 2 seconds at a time, the effect would be even more pronounced....

One thing that came to my mind yesterday is that of the effect of interlacing. Unless you have a progressive scan camcorder, each frame will be recorded as the usual two fields. Even with true single frame sampling for time lapse, the field sequence will not be correct. Strictly, new interlaced frames should be generated.